The Computer Working Group of the International Association of Egyptologists has been in existence since 1983. The group
147 101 72MB
English Pages 228 [224] Year 2009
Information Technology and Egyptology in 2008
Bible in Technology Volume 2
Series Editor Keith H. Reeves
Information Technology and Egyptology in 2008 Proceedings of the meeting of the Computer Working Group of the International Association of Egyptologists (Informatique et Egyptologie), Vienna, 8–11 July 2008
Edited by Nigel Strudwick
Gorgias Press 2008
First Gorgias Press Edition, 2008 Copyright © 2008 by Gorgias Press LLC All rights reserved under International and Pan-American Copyright Conventions. No part of this publication may be reproduced, stored in a retrieval system or transmitted in any form or by any means, electronic, mechanical, photocopying, recording, scanning or otherwise without the prior written permission of Gorgias Press LLC. Published in the United States of America by Gorgias Press LLC, New Jersey ISBN 978-1-60724-068-6 ISSN 1943-9369
Gorgias Press
180 Centennial Ave., Suite 3, Piscataway, NJ 08854 USA www.gorgiaspress.com Library of Congress Cataloging-in-Publication Data IAE Computer Working Group. Meeting (2008 : Vienna, Austria) Information technology and Egyptology in 2008 : proceedings of the Meeting of the Computer Working Group of the International Association of Egyptologists (Informatique et Egyptologie), Vienna, 8-11 July, 2008 / edited by Nigel Strudwick. -- 1st Gorgias Press ed. p. cm. -- (Bible in technology ; 2) 1. Egyptology--Information technology--Congresses. 2. Egyptology--Data processing--Congresses. I. Strudwick, Nigel. II. Title. DT60.I34 2008 025.06’932--dc22 2008051347 The paper used in this publication meets the minimum requirements of the American National Standards. Printed in the United States of America
CONTENTS
INTRODUCTION
Nigel Strudwick . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 SECTION I PAPERS FROM THE VIENNA MEETING THE ROSETTE PROJECT: COMPUTER ASSISTANCE FOR THE STUDENT, THE EPIGRAPHIST AND THE PHILOLOGIST
Vincent Euverte . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9 TRISMEGISTOS. AN INTERDISCIPLINARY PORTAL OF PAPYROLOGICAL AND EPIGRAPHICAL RESOURCES
Svenja A. Gülden . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17 1000–500 BC Claus Jurman, University of Vienna . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29 THE MEMPHIS DATABASE PROJECT
EDUCATIONAL IMAGES ON THE WEB
Edward Loring . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45 DAS GEFLÜGELTE KROKODIL CODIERUNG VON TOTENBUCH-VIGNETTEN
Marcus Müller-Roth . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49 AUTOMATIC ALIGNMENT OF HIEROGLYPHS AND TRANSLITERATION
Mark-Jan Nederhof . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71 CORPUS ÉLECTRONIQUES DE L’ANCIEN ÉGYPTIEN: TRAITEMENT XML DES TEXTES DES PROCESSIONS DE SOUBASSEMENT DES TEMPLES TARDIFS
Vincent Razanajao . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
V
VI
INFORMATION TECHNOLOGY AND EGYPTOLOGY
2008
AN OFFERING TO AMUN-RA: BUILDING A VIRTUAL REALITY MODEL OF KARNAK
Elaine Sullivan, Willeke Wendrich . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109 AGENT-BASED MODELS OF ANCIENT EGYPT Sarah Symons & Derek Raine . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 129 L’USAGE DE LA 3D EN ARCHÉOLOGIE
Robert Vergnieux . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 147 RAMSES. A NEW RESEARCH TOOL IN PHILOLOGY AND LINGUISTICS
S. Rosmorduc, St. Polis, J. Winand. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 155 SECTION II ADDITIONAL PAPERS AUTOMATED TRANSLITERATION OF EGYPTIAN HIEROGLYPHS
Serge Rosmorduc . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 167 RELATIONAL DATABASE DESIGN: A TUTORIAL AND CASE STUDY FOR EGYPTOLOGISTS
Ernest W. Adams and Nigel Strudwick . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 183 USING RELATIONAL DATABASES AT THE BEGINNING OF THE CENTURY
21ST
Dag Bergman . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 209 Author Addresses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 221
INTRODUCTION
Nigel Strudwick
At the closing session of the 2006 Oxford meeting of the Computer Working Group of the International Association of Egyptologists (Informatique et Egyptologie, I&E) an invitation was extended to the group by Regina Hölzl to hold the next meeting at the Kunsthistorisches Museum in Vienna (KHM) in 2008. Those present accepted this offer with alacrity, and the conference of which the present volume is the proceedings is the result. I should, on behalf of the group, like to thank Dr Wilfried Seipel, Director of the KHM, for his generosity and hospitality in agreeing that the meeting could take place in his museum. The organisation of the meeting was undertaken by Regina Hölzl and her colleagues in the Egyptian Department of the KHM, and I and all the participants are deeply appreciative of the welcome and excellent facilities extended to us. BACKGROUND TO THE PRESENT PUBLICATION
Throughout the 1980s and into the early 1990s, it proved possible to publish the proceedings of the meetings in print through the good offices of founding members of the group Nicolas Grimal and Dirk van der Plas. The last proceedings to appear in this way were those of the 1994 meeting in Bordeaux. The publication of subsequent meetings has been erratic, with some papers appearing independently, and only the papers from the 2002 Pisa meeting saw the systematic light of day via a CD ROM. Publication via electronic media seems very appropriate for a group whose interests lie in the possibilities of Information Technology in Egyptology. Nonetheless, it cannot be denied that there is still considerable 1
2
INFORMATION TECHNOLOGY AND EGYPTOLOGY
2008
prejudice against, or suspicion of, such publications within the subject. The manner in which such works are collected and catalogued within Egyptology is still a little unreliable, and the rather fragile aspects of the Internet, particularly the ease with which sites come and go, is still very much an issue. For whatever reason, a print publication still seems more likely to come to the attention to colleagues, and has potential advantages in raising the profile of I&E. Thus I was delighted to be approached by Katie Stott, Production Editor of Gorgias Press, with the offer of publishing these papers. I am very grateful to Gorgias for taking on this task. It is likely that, with the agreement of the publishers, some of the papers in this volume may also appear electronically on the web-sites of individual contributors. CONTENTS OF THIS VOLUME
A total of sixteen papers was presented at the meeting, arranged into four broad groups. The first, Modelling and Animation, produced an excellent introduction by Robert Vergnieux to the processes and issues of modelling in 3D. The world of 3D has much to offer Egyptologists in terms of reconstructing ancient worlds, and a particular example of it was presented by Elaine Sullivan with a very informative and elaborate model of the temples of Karnak. The paper of Raine and Symons introduced to I&E the subject of Complex Systems, and the modelling or simulation of social processes in ancient societies. Such modelling may seem doomed to failure in ancient Egypt due to the very skewed nature of surviving data, but careful application of the principles in conjunction with full understanding of the data can produce interesting results. Most importantly of all, it brings further techniques to the attention of Egyptologists. Session 2 was devoted to Text Corpora and Text Processing. Papers by Razanajao and Grützkau presented examples of the use of XML as an open standard for the encoding of texts. The implementation of an ambitious database of Late Egyptian texts was described by Rosmorduc and Winand, a database which has a real potential as a research tool for many users. Michael Everson’s paper gave a brief update on the status of the proposal for the encoding of hieroglyphs in Unicode, asked several questions of the group, and made a number of suggestions for the future. Nederhof ’s paper examined options for the alignment of the hieroglyphic and transliterated versions of texts, and illustrates how techniques of analysis from Computer Science can provide useful features and insights for Egyptologists.
INTRODUCTION
3
Databases have for many years been a central feature of presentations at I&E, and new and old ones featured in Session 3. As new databases, Navrátilová & Landgrafová looked at the IT dimensions of their database of First Intermediate Period Biographical Texts, as Jurman did with his database of Late Period material from Memphis. The discussions of these papers revolved around a number of points, but particularly how such datasets might be released onto the Internet and made available for all, an issue which needs to concern I&E very much in the future. Two projects with prominent presences on the World-Wide Web completed this session, with Gülden discussing the papyrological collections of Trismegistos, and illustrating how it works, while Horst Beinlich demonstrated the SERat database at the University of Würzburg, and showed a number of features of the project which thus far have not been implemented in the online version, but which may be consulted via the department in Würzburg. The last session was entitled Images, Bibliography and Tools. As for the first of these, Müller discussed how the complex vignettes of the Book of the Dead may be registered, while Loring presented his thoughts on the online availability of images and issues of museum copyright. The final paper was an update from Willem Hovestreydt on the Annual Egyptological Bibliography, to which I will return shortly. The opportunity has been taken to include two papers not presented at the Vienna meeting. That by Serge Rosmorduc was presented originally at the Würzburg I&E conference in 2000, but never published. Since that time his system has developed, and it is highly appropriate to present it here now as a up-to-date examination of how computers might be used to begin to attempt automatic analysis of hieroglyphic texts, particularly in the context of the paper by Nederhof which appears here. The second paper is a republication of a paper on database theory by Ernest Adams and myself presented in 1986 and published in 1990. A paper kindly contributed by Dag Bergman follows it here to help explain its inclusion and relevance. The technology may have changed but the underlying principles have not, and it would seem that although there are many Egyptologists using databases, a considerable number of them do not fully understand the underpinnings of the systems they use. As the original publication has long been unavailable, I decided to take advantage of the present volume to make it available again. For simplicity of setting, all references in papers have been left in the formats in which they were submitted by the authors.
4
INFORMATION TECHNOLOGY AND EGYPTOLOGY
2008
CONCLUDING REMARKS AND THE FUTURE OF I&E
By way of closing remarks at the Vienna meeting, I reviewed the subjects discussed, indicated any actions which the meeting should take, and looked into the future. Two business matters should be recorded here. Michael Everson had asked in his paper for the group’s opinion on the best way of encoding the transliterated Egyptological yodh, which apparently does not exist yet within any Unicode font. Summarising, the options were to encode it as a discrete character, or to create it from existing elements in other Unicode fonts. The meeting felt that the former course was better, recognising that it would be a longer process than creating a composite; one important reason for not preferring the composite was that it would free us from any obligation to have another font on the computer which would remain otherwise unused. Everson also enquired about the default sort order for the signs in the Unicode encoding of hieroglyphs; as there is no consistent or overwhelming way to handle all of them phonetically, it was felt best that the sort order should be that of the Gardiner list. It was stressed that I&E does not feel it has the right on its own to make such decisions on the part of the whole community, but I passed on its recommendations to the President of the International Association of Egyptologists (IAE), James P. Allen. The response on the issue of yodh from the Unicode Technical Committee (UTC) which oversees Unicode proposals, has thus far not been encouraging. They seem set against the creation of new characters, preferring composites. I am in the process of making representations against this, although I am not optimistic that the views of those who actually use the characters are of particular significance to the UTC. The precarious situation of the Annual Egyptological Bibliography has long been a cause for concern in the Egyptology world. Since Vienna, major developments have taken place, and Willem Hovestreydt has authorised me to print this statement on the state of matters in January 2009: At the Vienna meeting I described the AEB’s financial and institutional situation, which had become very difficult, especially after the University of Leiden decided to cease funding by the end of 2008. I presented several options, one of which was that the AEB should seriously explore the possibility of moving its base to a foreign institution. I am happy to say now that such a solution has indeed been found. From 1 January 2009 the AEB will be located
INTRODUCTION
5
in the Griffith Institute, Faculty of Oriental Studies, University of Oxford. The name Annual Egyptological Bibliography will change to Online Egyptological Bibliography (OEB). The OEB will also include Christine Beinlich-Seeber’s Bibliographie Altägypten 1822–1946. Discussions begun in Vienna played an important role in generating the initiative for the move to Oxford. I am sure all Egyptologists, not just participants in I&E, will wish the AEB the very best of futures, and will continue to support the project. I was delighted to be able to announce that Jean Winand, one of only two members of the group to be present at the first meeting of I&E (the other is Robert Vergnieux), has offered the University of Liège as the venue for the 2010 meeting of the group. All members were delighted at his generosity, and we look forward to reconvening in Belgium. The issue of the future of I&E frequently crosses my mind. In the early days of the group, it produced work widely used by the rest of the Egyptological community, notably the Manuel de Codage system, and the Multilingual Egyptological Thesaurus, as well as watching over the production of the earliest systems for writing hieroglyphs with computers. The most important thing which I&E should do is to continue to make itself relevant to Egyptology, and encourage those who are using various innovative IT techniques to bring them to meetings so that they may become wider known and benefit from discussions with others. One such area which is becoming more frequent in Egyptology, but which has yet to make an appearance at I&E, is Geographical Information Systems (GIS). But there are many areas which the meeting has examined before and which are as important as ever. I have already noted that fuller understanding of database concepts needs reinforcement and reiteration. Another matter which I&E has discussed and which needs further examination is that of data persistence and preservation (see my comments earlier about electronic vs print publication!). It is my belief that in this latter area I&E can help drive forward systematic approaches, and to that end, it was agreed in the closing session of the meeting that those of us interested in this topic should meet in the course of 2009 and consider how discussion on this subject should be facilitated at the Liège meeting. The future for I&E remains exciting so long as we remain relevant and encourage our colleagues to take full advantage of the information and technological revolutions which we have observed and participated in over the past 25 years.
6
INFORMATION TECHNOLOGY AND EGYPTOLOGY
2008
PAPERS PRESENTED AT THE VIENNA MEETING
Session 1 (Modelling and Animation) Vergnieux, L’usage de la 3D en Egyptologie Sullivan, An Offering to Amun-Ra: Building a Virtual Reality Model of KarnakTemple Raine & Symons, Complex Systems: Agent based models of ancient Egypt Session 2 (Text Corpora and Text Processing) Razanajao, The Electronic Text Corpora of Ancient Egyptian: XML Treatment of Processions of Nile-Gods Grützkau, Hieroglyphic Texts and XML* Rosmorduc & Winand, Ramsès Project Everson, Yod, Unicode, and future options for Egyptian encoding* Nederhof, Automatic alignment of hieroglyphic and transliteration Session 3 (Databases) Navrátilová & Landgrafová, The Database of First Intermediate Period Biographical Texts* Jurman, Prosopography and Monument-analysis with the Memphite Late Period Database Beinlich, Datenrecherche über die Internet-Version von SERat hinaus* Gülden, Trismegistos Session 4 (Images, Bibliography and Tools) Müller, Encoding Vignettes of the Book of the Dead Loring, Sharing Images in the Internet Euverte, Rosette. A computer-assistance for the student, the epigraphist, and the philologist Hovestreydt, The Annual Egyptological Bibliography: Recent and Latest Developments* Nigel Strudwick, Closing comments * indicates that this paper was not submitted for this volume.
THE ROSETTE PROJECT: COMPUTER ASSISTANCE FOR THE STUDENT, THE EPIGRAPHIST AND THE PHILOLOGIST
Vincent Euverte
ABSTRACT
Since 2004, the Rosette Project has promoted the culture of Ancient Egypt through its writing, with its main focus on the usage of modern technologies to assist in the reading and the translation of hieroglyphs. This paper describes the achievements over these four years, and highlights the perspectives offered by the rapid developments in computer science. INTRODUCTION
In 2004, the concept of computer assistance to read and translate the hieroglyphic texts was launched. Soon some Egyptology amateurs joined the project, and the next two years were dedicated to address multiple difficulties in programming as well as in the learning of the writing. In July 2006, the Informatique & Egyptologie Computer Group Meeting in Oxford gave the first opportunity to present our initial results to professional colleagues. The reception given to the project was encouraging, and many pertinent items of advice were offered so as to orientate future developments. In 2008, two major events have allowed us to reassess the project and to redefine our axes of progress: the International Congress of Egyptology in Rhodes, and the Informatique & Egyptologie Computer Group Meeting in Vienna. 9
10
INFORMATION TECHNOLOGY AND EGYPTOLOGY
Figure 1: The home page of the Rosette Project
Figure 2: Sample entry from the catalogue of hieroglyphs
2008
THE ROSETTE PROJECT
11
We wish here to thank all those who believed in and supported the Rosette project. Their suggestions and comments will be incorporated into the future development of the project. THE ROSETTE PROJECT
Thanks to a team of committed and talented volunteers, and supported by a network of hundreds of users, the Rosette application (http://projetrosette.info/, Figure 1) has been on-line and free of charge on the Internet for three years and offers the following features: • the multilingual environment includes so far French, English, and Arabic (Spanish and Portuguese are under development); a simple and dynamic interface with drop-down menus and an integrated help function; compatibility is assured for most web browsers which are in compliance with XHTML standards, without any influence on the user environment (there is no need for cookies nor downloads). • The catalogue of hieroglyphs (Figure 2) contains more than 3000 signs, documented and enriched with palaeographic images, and available in three different fonts in Unicode EGPZ format, in partnership with Saqqara Technology. A descriptive card is accessible by a single click on any hieroglyph, from everywhere in the Rosette application. • The dictionary already contains all the entries of the Concise Dictionary of Middle Egyptian by R.O. Faulkner, with the kind authorisation of the Griffith Institute, as well as its French counterpart from the Medjat Association. Multi-criteria search tools allow access to different writing forms, semantic variants, transliterations, translations, and bibliographic references. • The hieroglyphic editor complies with the Manuel de Codage 1988 (MdC) and offers several additional functions such as the fine positioning of the signs and the composition of groups and sign combinations. A lexical analysis allows to find automatically every word of the dictionary that is contained in the submitted sentence. • A corpus of texts links high resolution photographs of the original artefact with the hieroglyphic pattern, the MdC sequence, the transliteration, the translation in several languages, and the references to the source document. Each artefact is described according to the
12
INFORMATION TECHNOLOGY AND EGYPTOLOGY
2008
Figure 3: sample text corpus entry, from the White Chapel of Senwosret I
Multilingual Egyptian Thesaurus (MET) to allow thematic searches. An ongoing project is an indicator of developments to come: a detailed visit to the White Chapel of Senwosret I in Karnak (Figure 3). 3D navigation will be available for every scene, and permit the analysis of the symmetrical arrangement of this wonderful monument. • Several thematic articles discuss epigraphy, philology, calligraphy, and so on, as well as the lists of kings, gods, toponyms, etc. in association with the dictionary. The tools we have developed enable a wide range of topics to be examined, and we are planning to expand the scope of this further. An interesting example under construction is the list of military titles at all dates, collected by Michel Sancho, a specialist in military matters. ASSESSMENT
The comments received from the professionals are encouraging to our approach. The following are the most appreciated strengths of the Rosette project: • A structured database: From its initial conception, Rosette has been seeking an information architecture which will lead to an
THE ROSETTE PROJECT
•
•
•
•
•
13
optimal usage. However, Egyptology has multiple needs which require reassessment and realignment of this structure. Dag Bergman and Nigel Strudwick demonstrated their interest in such a review. Therefore, we will document our approach and submit it to them for their review and suggestions. Integration: One of the key features of Rosette is its ability to integrate very varied functions in a single environment. Most of these functions already exist in other application, often more powerful overall, but rarely combined altogether. The interaction between the catalogue, the dictionary, the editor, and the corpus makes Rosette a global tool for the study of hieroglyphs. Referees: Every single element (MdC, transliteration, translation, bibliographic reference) has to be checked by a qualified person. Our approach is to grant a ‘referee level’ to the reviewer, and to indicate the review date, the name and the level (at this date) of the reviewer. Attestations: Rosette aims to explain each source as much as possible. A good example is the palaeographic collection, which justifies the choice we made in the design of our hieroglyphic fonts. Respect of copyright: The Rosette project is proud to be assisted by an international Law Advisory Agency (Hogan & Hartson). We took the firm resolution to protect the authors’ rights, and to only publish duly authorised works. We believe this to be an assurance of our seriousness and our intentions to maintain this project over the longer term. Compliance with the standards in force in Egyptology: The Manuel de Codage, the MET, the basics of transliteration, the bibliographic abbreviations, and so on, are all essential to the acceptance of the project by specialists as well as by amateurs.
PERSPECTIVES
The Rosette Project has already achieved number of its initial objectives. Nevertheless, we are far from having exhausted all the possibilities offered by the technology. So it is even more important to highlight the major axes of improvement: • Harmonisation and extension of the Rosette fonts: Whatever its quality, a computer font is reductive by nature, for it normalises
14
INFORMATION TECHNOLOGY AND EGYPTOLOGY
•
•
•
•
2008
the graphic representation of each hieroglyph. So it is worth debating whether it is necessary to build a collection of fonts covering each period and/or each writing mode. One could for instance envisage a first series of five fonts to handle separately Middle Egyptian, Late Egyptian, Late Period Egyptian, a ‘printing font’ (e.g. de Rougé or Theinardt), and a semi-cursive one for papyrus texts. Rosette already proposes the first three and is evaluating the others. Revision and extension of the dictionary: First, it is essential to validate our current database. This effort requires skills that are beyond our team’s resources as of today, and it stresses the need for the project to be supported by philologists. As it is currently structured, the Rosette dictionary is ready to gather multiple sources. The collections of royal names, divine names, military titles, toponyms, and so on, are underway; there are many thematic items which will enrich the dictionary and the lexical analysis. Enrichment of the corpus and cross-analyses: The Rosette corpus currently includes nearly one hundred texts. In order to achieve cross-analyses on different criteria (period, supporting material, type of text, and so on), this collection must grow. The MET classification is a first step towards rationalisation. We are now developing multi-criteria search tools and adapting the existing texts to these standards. Another axis of progress is the statistical analysis of the occurrences of hieroglyphs by period, supporting material, and so on. On another hand, we also have numerous photographs of original artefacts readily available for future addition to the corpus. As soon as volunteers are available, their translations can be immediately verified and posted on the web-site. Semantic and lexicographic analyses: Transliteration is unavoidable in the translation process. With the help of the catalogue, the dictionary, and the phonetic table, it should be possible to facilitate and to speed-up this exercise with the assistance of the computer. Then, a semi-automatic analysis should allow the identification of a few elementary grammatical rules to facilitate the work of the reader-translator. We are still aware that his/her acumen and experience will be required for the final interpretation. Thematic articles on epigraphy and philology: The Rosette Project’s purpose claims to ‘promote the culture of Ancient Egypt through its writing’. The corollary of this wish is to publish serious
THE ROSETTE PROJECT
15
and documented articles about any topic related to hieroglyphic writing. • These are not withstanding the many potential improvements of the web-site itself, to ease navigation for the visitor. CONCLUSIONS
Computer science offers almost boundless opportunities. It is up to us to exploit them for the sake of the Egyptological community. Should a worldwide standard emerge in the coming years, it could allow all texts to be encoded in a unique format, available in one single database, with a common dictionary encompassing every expert’s knowledge, assisted by a syntactic engine, and in multi-lingual translations. Even if these standards are not yet ready, nothing prevents starting to capture the data in a temporary format. Whatever Rosette can collect from now on, it will be easy to convert and transfer it to any future ‘official database’, blessed by the Egyptological community, as soon as this database is created. Is not it worth starting the data collection now, in parallel with the development of the new standards?
TRISMEGISTOS. AN INTERDISCIPLINARY PORTAL OF PAPYROLOGICAL AND EPIGRAPHICAL RESOURCES
Svenja A. Gülden
ABSTRACT
Trismegistos, named after the famous epithet of Hermes-Thoth, the Egyptian god of wisdom and writing who also played a major role in Greek religion and philosophy, is a platform aiming to surmount barriers of language and discipline in the study of late period Egypt and the Nile valley (roughly 800 BC–AD 800). It brings together a variety of projects dealing with metadata, mainly of published documents. Its core component is Trismegistos Texts, which includes papyrological and epigraphic texts, not only in Greek, Latin, and Egyptian in its various scripts (Demotic, Hieroglyphic, Hieratic and Coptic), but also in Meroitic, Aramaic, Arabic, Nabataean, Carian, and other languages (currently 106859 records).
Scholarly tools that help to find and explore written sources from Ancient Egypt with relevance for one’s research have become increasingly important in the last century. For Demotic studies such tools were first provided in the late 1960s and the early 1980s, mainly by articles of Pestman, Thissen and Lüddeckens, that made precisely dated demotic texts easily retrievable through detailed lists.1 For Greek papyrology such a tool was missing at that time. Since the early nineties of the previous century, however, the Heidelberger Gesamtverzeichnis der griechischen Papyrusurkunden Ägyptens (HGV) provided similar, but digital lists of dated documents, and from 1997 onwards their FileMaker 17
18
INFORMATION TECHNOLOGY AND EGYPTOLOGY
2008
database with information about date, provenance and other metadata has become fully accessible online.2 For Greek literary texts a similar tool is available in the form of the Leuven Database of Ancient Books (LDAB).3 Digital tools like these of course made metadata searches more flexible for Greek texts than that for Demotic ones. Therefore the first goal of the project Multilingualism and Multiculturalism in Graeco Roman Egypt4 was to create a similar tool to find and explore Demotic and other Egyptian papyrological material in general. SETTING UP TRISMEGISTOS
In early 2005 we therefore started to create Egyptological metadatabases as counterparts to the already mentioned databases HGV – concentrating on Greek and Latin documentary papyrological texts – and LDAB – for literary texts – by programming the databases DAHT (Demotic and Abnormal Hieratic Texts) and HHP (Hieroglyphic and Hieratic Papyri). Fortunately, for neither of these projects (DAHT and HHP) did we have to start from nothing, because for Demotic papyrological material two unpublished digital tools were placed at our disposal. The first was a database of limited metadata of Demotic papyri of all genres, compiled by Heinz-Josef Thissen, and the second was the nascent database with more elaborate metadata of Demotic documentary texts (written on papyri, ostraca and occasional other writing surfaces), which was part of the relational databases of the Prosopographia Ptolemaica and the Leuven Homepage of Papyrus Collections.5 1
2 3 4
5
P.W. Pestman, Chronologie égyptienne d’après les textes démotiques, P.L.Bat. 15, Leiden 1967, H.-J. Thissen, Chronologie der frühdemotischen Papyri, in: Enchoria 10, 1980, 105–25, E. LÜDDECKENS, Papyri, Demotische, in: LÄ IV, Wiesbaden 1982, 750– 898. http://aquila.papy.uni-heidelberg.de/gvzFM.html http:// ldab.arts.kuleuven.be/ This project was made possible by a Sofja Kovalevskaja Preis awarded by the Alexander von Humboldt-Stiftung. The award which was obtained by Mark Depauw from the Katholieke Universiteit Leuven did allow him to set up this project at the Seminar für Ägyptologie der Universität zu Köln, under the auspices of Heinz-Josef Thissen, the former Director of the Cologne Egyptological Seminar. http://lhpc.arts.kuleuven.ac.be/
TRISMEGISTOS
19
After these two unpublished databases had been merged into the new system and double entries had been eliminated, new records were entered on the basis of the Berichtigungsliste of Demotic Documents6 – a non-digital tool listing all published Demotic texts for which corrections had been proposed in secondary literature. The HHP database also builds upon two unpublished digital tools which were put at the project’s disposal: first the database with metadata of Late Period Funerary Papyri created by Marc Coenen and second that of Late Book of the Dead Papyri compiled mainly by Ursula Verhoeven with later additions by the present author. The second database was adapted and enlarged by the Book of the Dead project in Cologne and Bonn and published as an Online Prosopographie in a downloadable version with limited data, which – at a later stage of my work on the HHP – also proved useful.7 Again eliminating double entries and assigning a unique number to the texts in the system was one of the main elements of the work, but in this case another, sometimes complicated task was identifying and joining papyrus fragments that belong together in one single entry. Each entry in our databases is considered to be a single document. To determine what constitutes a ‘single document’ we have adopted a definition based on material aspects: In principle all texts written on what was in antiquity a single writing surface belong together and form one document, unless there are good reasons to believe that the only (and unrelated) relation between the two texts is the writing surface itself.8 If two or more texts on the same writing surface are related, they are therefore considered a single document; if the texts do not belong together, i.e. in case of reuse, they form separate entries in the database but are connected by a link to each other. Thus papyrus fragments that are spread over several museums (e.g. funerary papyri such as Book of the Dead manuscripts) but which can be identified as being parts of a single papyrus in 6
7 8
A.A. Den Brinker, B.P. Muhs, S.P. Vleeming, A Berichtigungsliste of Demotic Documents, Studia Demotica 7, 2 vols, Leuven 2005. Although itself at the time unpublished, the compiler Sven Vleeming was so kind as to send a preliminary version at the start of the present project, which greatly facilitated data entry. http://www.uni-bonn.de/www/Totenbuch_Projekt/Online_Prosopographie.html See M. Depauw, Bilingual Greek – Demotic Documentary Papyri and Hellenization in Ptolemaic Egypt. The Trismegistos Database, in: P. Van Nuffelen, (ed.), Faces of Hellenism, Studia Hellenistica, Leuven, forthcoming.
20
INFORMATION TECHNOLOGY AND EGYPTOLOGY
2008
antiquity, constitute a single record. This basic definition works quite well for the papyrological material, although there are, as always, some problematic cases. It soon became evident that the two language / script based databases DAHT and HHP would hardly be practicable without an accompanying bibliographical database. Therefore we started with the digital version of the Demotistische Literaturübersicht (DL). This bibliographical tool was initially restricted to the entries made in the published version in Enchoria since 1971. We again were fortunate to receive a basic electronic version from H.-J. Thissen. In addition to this we have created a general bibliographic database (TMBib), which is linked directly to our text databases. It provides bibliographical information for all DAHT and HHP entries, including the abbreviations used. Since this bibliography is fully implemented in the system it is also possible to start from a certain publication and see whether there are links to the texts themselves. Apart from setting up databases for material in Egyptian scripts as counterparts to the Greek papyrological ones, the project also wanted to investigate questions of multilingualism on the basis of these tools. To quantify we therefore needed to establish ways of exchanging information by bringing everything together in a common platform, which we have decided to call Trismegistos. For this purpose we first implemented a mapping procedure for all records with metadata of the HGV, LDAB and our several project databases. Each record of the partner projects was then assigned a unique (and random) identification number, the TM-ID, which is used to link between all relevant databases. TRISMEGISTOS NOW
– OFFLINE VERSION
Trismegistos is based upon several internal and external partner projects: Internal Trismegistos partner projects DAHT Demotic and Abnormal Hieratic Texts HHP/HHT Hieroglyphic and Hieratic Papyri / Texts ATE Aramaic Texts from Egypt TM Magic Magical, ritual, religious and divinatory Texts (all languages) LDAB Leuven Database of Ancient Books (all languages)
TRISMEGISTOS
21
External Trismegistos partner projects HGV Heidelberger Gesamtverzeichnis (Greek and Latin) APD Arabic Papyrology Database BCD Brussels Coptic Database
All Trismegistos partner projects9 collect information in a shared offline FileMaker server database system.10 It consists of a set of about 16 (main) relational databases with metadata. The core are over 105,000 text or document records, linked to this are about 157,000 publication entries and just over 109,000 dates. Together with the Leuven Homepage of Papyrus Collections, these documents are linked to more than 100,000 inventory numbers and about 1,400 collections; in collaboration with the Fayum project (Leuven) the more than 117,000 provenances of these documents are linked to about 7,600 places. Finally 375 archives – collected by the Leuven Archives project – could be connected in 12,600 cases to our entries. As a result of the close cooperation of all Trismegistos partner projects the database includes metadata of various kinds and scripts of written material (Table 1). TRISMEGISTOS NOW
– ONLINE VERSION
Another important aim was not only to set up this database system for our own research concerning multilingualism and language shifts in GraecoRoman Egypt, but also to share information and make our database available to all interested scholars by creating a freely accessible online platform. Based on PHP / MySQL11 Trismegistos went online in a first version in November 2006 with only limited search facilities such as publication numbers and inventory numbers. The latest version of Trismegistos went online in February 2008 at http://www.trismegistos.org/. The six main modules (Figure 1) of Trismegistos are: Texts, Collections, Archives, People, Places and Bibliography, whereas the core component is Trismegistos Texts (Figure 2).
9 10 11
HGV, APD and BCD provide us with regular updates so that we can include actual information in Trismegistos. The database structure was designed mainly by Bart Van Beek and the present author in an FileMaker Pro 7 / 8 / 9 environment. The online version was designed by Jeroen Clarysse and Bart Van Beek.
22
INFORMATION TECHNOLOGY AND EGYPTOLOGY
Greek & Latin Demotic & Abnormal Hieratic Coptic Hieroglyphic & Hieratic Meroitic Aramaic Arabic Other languages & scripts
2008
Papyrological Epigraphical Documentary Literary HGV LDAB Gr.-Lat. IGLE 52400 records 10436 records ~ 10000 records DAHT (incl. LDAB Dem.) 13975 records (incl. 654 LDAB Dem.) BCD LDAB Copt. no partner 7074 records 1590 records 1833 records HHP no partner 1509 records 3452 records no partner 940 records ATE 1150 records APD no partner 566 records no records no partner ca. 400 records
Table 1: Types of metadata in Trismegistos by types and partner sources
Figure 1: Trismegistos online portal
TRISMEGISTOS
23
Figure 2: Trismegistos Partner Projects Texts databases
In addition to the already mentioned partner projects, the CPP (Corpus of Paraliterary Papyri) is listed (at the very right), which does not provide any information but does link directly to Trismegistos Texts by using the TMnumber. More information about all of our partner projects, as well as links to their home pages, can be obtained by using the corresponding button. The new search form provides more elaborate search fields than we had in our first online version: Publication, Editor, Inventory number, Material, Language/Script, Provenance, Nome/Region and Date.12 Search criteria can be combined and every search field allows two requests by the use of the Boolean ‘OR’, ‘AND’ or ‘BUT NOT’. A short explanation accompanies every search field – for example to define the ‘strict’ button that filters the queried range of dates. The list of matching data gives the publication number(s) and the inventory number(s). To save a certain search result from Trismegistos, a 12
For dates BC, negative numbers are used. The dates in the document sheets, however, are shown in the common way using BC and AD.
24
INFORMATION TECHNOLOGY AND EGYPTOLOGY
2008
Figure 3: DAHT – Demotic and Abnormal Hieratic Texts
button ‘export to .csv’ (still under development) at the top of all lists permits the download of the search results into a format which can be imported, for example, into an Excel spreadsheet. For more details about a particular record two possibilities are given: The TM ID – Link for limited metadata or the link to the partner project (e.g. the databases HHP, DAHT or HGV)13 with more elaborate information such as the type of text, details concerning the language, the use of recto and verso, the provenance, the date and so on (Figure 3).14
13 14
In those cases where a text is found in more than one partner projects all possible links are given (e.g. HGV and DAHT). The elaborate data sheets provide not only actual inventory numbers but also show older numbers that are out of use and even wrong inventory numbers used in literature. As well as starting with a general search in Trismegistos, it is possible to look up a special subset of data by using one of the partner databases that leads directly to the elaborate data sheets.
TRISMEGISTOS
25
Figure 4: Collections – LHPC data sheet
In some cases, additional literature – with key words concerning the topic of the relevant entry – is listed below the data sheet. This bibliography does not claim to be exhaustive – it is just an extract of recent or particularly relevant literature. By using the links (indicated by underscore) the user can not only obtain the full bibliographic data but also a list of related and linked texts in Trismegistos. At the left side more related data – collections info and archives info – is listed and of course linked to the respective database. The link to collections leads to a data sheet with limited data about the collection itself but it provides a list of other Trismegistos texts present there. Through cooperation with the Leuven Homepage of Papyrus Collections (LHPC) project, more detailed information concerning the collection is often provided on the LHPC data sheet (Figure 4). The archives-info link – provided in cases where a text is known to belong to a certain archive in antiquity – leads to a data sheet with detailed
26
INFORMATION TECHNOLOGY AND EGYPTOLOGY
2008
Figure 5: Archives – data sheet
information about the relevant archive (Figure 5). In this case the infobox at the left shows the publication numbers of all Trismegistos texts linked to this archive. Whenever possible we try to give information concerning the provenance of a document, where it was found, written or, in the case of letters, what its probable destination was. The link provided leads the user to the places database that shows – because the example above chosen was Memphis – not only Memphis but also other nearby sites that are included under this entry. The most important variants are named as well as e.g. Greek, Demotic or Latin versions. Apart from links to Trismegistos texts that are linked with this area, a new feature, a link to Google Maps, is added. The red rectangles (grey in this figure) show places that are linked with Trismegistos (Figure 6). This is a very interesting feature, particularly for lesserknown places. Furthermore it is now possible to browse Google Maps for ancient place names in Egypt with the possibility to switch from Google
TRISMEGISTOS
27
Figure 6: Trismegistos and Google Maps
Maps directly to the Trismegistos places database. Of course place names in the Collections database are also linked to Google Maps. Starting from the core of Trismegistos, the Texts database, the main features of Trismegistos online with the databases Collections, Archives, Places and Bibliography have now been introduced here. But it is also possible to explore Trismegistos by using one of the other databases as a starting point. The people database, which was not described here, is still “under construction”. The project Multilingualism and Multiculturalism in Graeco Roman Egypt began work on the people database, but this work is now continued by the project Creating Identities in Graeco-Roman Egypt, led by Mark Depauw at the University of Leuven. Eventually all people mentioned in the Trismegistos texts should be included in this database, but in view of the very high number of individuals therein, cooperation with full-text databases, especially for Greek, is planned. Finally one more feature provided by Trismegistos should be mentioned: the Trismegistos Online Publications. This is a new series of freely downloadable PDF documents with scholarly tools based on the Trismegistos database. The first published volume is a Chronological Survey of Precisely
28
INFORMATION TECHNOLOGY AND EGYPTOLOGY
2008
Dated Demotic and Abnormal Hieratic Sources.15 Although this PDF of course can be printed, however, it should preferably used as a digital tool, because the entries provide a direct link to the Trismegistos online databases. The publication of a second volume listing placenames is imminent (October 2008). Trismegistos is still a ‘work in progress’. Despite the fact that there are without doubt missing and erroneous entries, we hope our current coverage is sufficient enough for Trismegistos to be a useful tool.
15
M. Depauw, C. Arlt, M. Elebaut, A. Georgila, S.A. Gülden, H. Knuf, J. Moje, F. Naether, H. Verreth, S. Bronischewski, B. Derichs, S. Eslah, M. Kromer, A Chronological Survey of Precisely Dated Demotic and Abnormal Hieratic Sources, Version 1.0 (February 2007), Köln / Leuven 2008, xiii, 232 pp: http://www.trismegistos.org/dl.php?id=4.
THE MEMPHIS DATABASE PROJECT
1000–500 BC
Claus Jurman, University of Vienna
ABSTRACT With the aim of creating a corpus of Third Intermediate Period and Late Period elite monuments from Memphis for prosopographical and sociological analysis, a flexible database capable of dealing with very different kinds of data becomes necessary. Since no commonly available ready-to-use Egyptological database software exists today, a custom-built database structure based on the common Microsoft Access application has been created for the purposes of an ongoing PhD project. Though being the product of a non-IT expert and tailored to the specific requirements of the study in question, the multi-purpose layout of the program, which covers among other things information on field archaeological contexts, dating, texts, palaeography, stylistic features, typology, personal names, titles, etc., could perhaps be successfully applied also by other Egyptologists conducting prosopographical or artefact-centred research.
INTRODUCTION TO THE PROJECT The Memphis Database Project is an outcome of my PhD dissertation entitled Memphis in der Dritten Zwischenzeit und der Spätzeit – (Selbst)Repräsentation auf Elitedenkmälern which is currently being prepared at the Institute of Egyptology, University of Vienna, under the supervision of Prof. Manfred Bietak. With the database still being a work in progress, this article cannot present definitive results. Instead it is meant to highlight certain aspects relevant to designing and working with an Egyptological database. The main goal of the aforementioned dissertation is to shed more light onto the history of Egypt’s traditional cultural capital and its citizens from c. 1000 to 500 BC by closely studying the elite monuments from votive and 29
30
INFORMATION TECHNOLOGY AND EGYPTOLOGY
2008
funerary contexts preserved at the site. Since comparatively little research has so far been directed to the region of Memphis between the end of the New Kingdom and the Persian Empire,1 it was decided to create at first a corpus of Memphite monuments from the Third Intermediate Period and the Late Period. This corpus covers objects that can be attributed to individuals by means of inscriptions or other criteria, dating from Dynasties 21 to 27. When completed, one of its primary assets will be that it assembles for the first time comprehensive sets of data on the totality of the relevant Memphite material now dispersed over unnumbered collections on several continents, including those objects that have not yet been properly published or indeed published at all. The following considerations had a decisive influence on the conception and realisation of the corpus by means of a relational database: • The primary research goal of the project is to investigate the changing modes of (self)representation of the Memphite elites on different kinds of monuments during the periods in question. In order to achieve this goal it is necessary to create a corpus of artefacts which can be linked to Memphite individuals through external (> e.g. archaeological records) or internal (> e.g. inscriptions) criteria. Additionally, a tool designed for this kind of research should facilitate creating dossiers for each recorded individual in which his or her monuments/attestations are grouped together according to pre-defined criteria of classification. • The individuals linked to the objects of investigation were subject to changing political and social conditions that should in one way or another have also found their expression in the objects themselves, namely through particular features of design or ‘decorum’.2 One has to account for the fact that the production of an artefact is a complex decision process, often taking place on more than one level of agency and involving deliberate choices as well as unconscious 1
2
Apart from excavation reports on the North Saqqara Temple Precincts and on the Late Period shaft tombs at Abusir, the most noteworthy publication of the last two decades touching upon the topic of the Memphis Database Project is C.M. ZIVIE-COCHE, Giza au premier millénaire. Autour du Temple d’Isis Dame des Pyramides, Boston 1991. On the concept of ‘decorum’ within Egyptology, see J. BAINES, Fecundity Figures. Egyptian Personification and the Iconology of a Genre, Warminster 1985, 278– 279.
THE MEMPHIS DATABASE PROJECT
1000–500 BC
31
ones. In order to filter out significant choices, a tight net of qualitative and quantitative parameters has to be spread over the object of study, providing the basis for a culturally meaningful ‘thick description’ in the sense of Ryles and Geertz.3 • Each object of research is conceived as a complex entity consisting of several partly interdependent cultural features and properties for which the term ‘culturemes’ will be used.4 Such culturemes may include, for example, particular iconographic modes of depiction (e.g. gestures of adoration), stylistic details (e.g. plastic rendering of eyebrows) or the particular type of stone used for an object, its proportions, measurements, types of texts present and conventions of spelling followed, and so on. • To make efficient use of the collected data, it is imperative to have a tool at hand which is capable of identifying patterns, i.e. testing potential connections between certain culturemes or groups of features. It might be relevant, for example, to investigate the systematic presence or absence of certain features (such as the patrilineal filiation in inscriptions identifying the owner of a monument) on particular groups of objects (such as funerary monuments).5 • In the fullness of time the corpus will comprise a great variety of objects, ranging from temple statues and funerary stelae to votive bronzes, funerary papyri, administrative documents, inscribed amphorae, etc. While the focus of research unquestionably rests on the abundant source material provided by statues and stelae, the corpus data will in the end include less prominent categories such as shabtis and votive figurines as well. 3
4
5
G. RYLE, Thinking of Thoughts. What Is ‘le Penseur’ Doing?, University of Saskatchewan University Lectures 18, 1968; see also http://lucy.ukc.ac.uk/CSACSIA/Vol14/Papers/ryle_1.html (last accessed on 14-09-2008); C. GEERTZ, Thick Description: Toward an Interpretive Theory of Culture, in: idem, The Interpretation of Cultures. Selected Essays, New York 1973, 3–30; esp. 5–6; 9–10. For the term ‘cultureme’, see I. EVEN-ZOHAR, Factors and Dependencies in Culture: A Revised Outline for Polysystem Culture Research, Canadian Review of Comparative Literature 24, 1997, 22. For the tendency to conceal the name of a person’s father within Memphite funerary monuments of the Late Period, see L. GESTERMANN, Die Überlieferung ausgewählter Texte altägyptischer Totenliteratur (‘Sargtexte’) in spätzeitlichen Grabanlagen, ÄA 68, Wiesbaden 2005, 402.
32
INFORMATION TECHNOLOGY AND EGYPTOLOGY
2008
• The corpus data should also be used for purposes of comparison with related material from other regions, for example, the group of statues from the first millennium BC found in the Karnak Cachette at Thebes.6 In Egyptological practice the two most common forms of data assemblage are the more or less detailed catalogue of objects with a fixed order following pre-defined categories, and the often alphabetically arranged catalogue of dossiers, which usually offers a basic selection of information on objects linked to particular individuals. In the first case, the range of objects considered is often rather limited. This makes it very difficult, if not impossible, to obtain an overview of a person’s activity at a site or in a region if this person is attested by more than one object category (e.g. by a funerary assemblage from the Saqqara necropolis, an Apis stela from the Serapeum at Saqqara and a votive statue from the temple of Ptah at Memphis). The second method of data processing has its shortcomings as well, since object-related information within a dossier entry is normally confined to the most basic level so as not to sacrifice clarity. Even if such an entry includes, for instance, the title sequence(s) and genealogical data provided by an inscribed monument, it will usually fall short of offering insight into the specific contexts in which they occur. However, these contexts are in many cases highly relevant for answering the type of questions sketched above. For example, it might be significant where a particular title occurs within a sequence or where a title sequence is positioned on an object. It could be important to know whether a divine name is part of an offering formula or appears within a biographical text as part of a priestly designation. Office-related titles may be found in separated groups or interspersed with epithets of a general nature. Certain types of genealogical data may only occur within specific contexts or in combination with particular iconographic attributes. However, it is indisputable that the registration and interpretation of these subtleties can hardly be addressed by the two methods delineated above. For the envisaged holistic approach of the Memphis Database Project, with its focus on the cultural and sociological significance of object-related properties, a different path had to be taken. As seen in Figure 1, the aim of an individual-centred monument analysis can only be achieved by 6
Cf. recently, J.-C. GOYON & C. CARDIN, Trèsors d’Égypte : la ‘cachette’ de Karnak, 1904–2004; exposition en hommage à George Legrain à l’occasion du IXe Congrès International des Égyptologues, Grenoble 2004.
THE MEMPHIS DATABASE PROJECT
1000–500 BC
33
Artefacts from a geographical entity (Memphis and its necropoleis in the early 1st millennium BC)
Corpus of monuments (catalogues)
Prosopography (dossiers)
Levels of agency (local tradition, commissioning person/ institution, workshop, artisan, etc.)
Individual-centred monument analysis
Figure 1: Scheme showing the major components of an individual-centred monument analysis
integrating the features of a detailed catalogue of objects and a collection of prosopographical dossiers. Such an integrated corpus as described in this paper should also help by enabling effective inner-corpus comparisons, and, in certain cases, by differentiating between several levels of agency contributing to the appearance of a monument. One might thus be able to filter out culturemes that are characteristic of a certain office or social position, of an owner/customer, a local workshop tradition or even an individual artisan. For all these reasons it appears as the logical choice to capitalise on the advantages of IT applications by creating a powerful and flexible database. DATABASE LAYOUT
– AIMS AND OBJECTIVES
Preceding the actual design of the database, a number of key-features were formulated without which the application would not meet the objectives. • The structure of the database should be flexible enough to ensure that it can be used, at least in principle, also within other areas of Egyptological research. By operating at a fundamental level with
34
INFORMATION TECHNOLOGY AND EGYPTOLOGY
2008
basic categories of object-related data that are not specific to the Memphite corpus, but of general relevance to museum collections and archaeological excavations, a high degree of data comparability would be guaranteed. This aim could be achieved by implementing different layers of detail within each entry form for an individual data category, and by generally making sure that the database layout remains adaptable and extensible. If these principles are adhered to, it should be fairly simple to integrate new object categories requiring their own entry form by reference in part to already existing data categories (e.g. for classifying pottery one would rather choose to enlarge the already existing hierarchically organised data category ‘material’ by several thesaurus entries than create a completely new category of ‘pottery fabrics’). • The user interface has to provide mechanisms through which the large amount of information collected for each object or individual is filtered so as to bring into focus the most important features. Otherwise one will risk forsaking the advantages that are connected with a comprehensive description. The solution might lie in providing the user with summary tabs in which all the names, divinities, titles, etc. occurring on a monument are listed together, ordered by quantity or relative importance. • Designed from the start as a multi-purpose application, the database should be able to deal with philological as well as with archaeological data. However, it is not the aim to provide a tool for full-scale linguistic analysis. With the Manuel de Codage (MdC)7 being the primary method for encoding Egyptian texts, the integration of hieroglyphs into the datasets is confined to jpeg-images (photos and/or images exported from common hieroglyphic editors such as Winglyph or JSesh8 and only undertaken if deemed appropriate. • In accordance with the principle of extensibility the database should offer possibilities for integrating specific ‘add-ins’ by way of hyperlinks. For example, elements of a flowchart created with Microsoft 7
8
J. BUURMAN et al., Inventaire des signes hiéroglyphiques en vue de leur saisie informatique, 3rd ed., Informatique & Egyptologie 2, Paris 1988. There is an unofficial web page by H. van den Berg dedicated to the Manuel de Codage: http:// www.catchpenny.org/codage/ (last accessed on 16-09-2008). For the latter, see http://jsesh.qenherkhopeshef.org/ (last accessed on 16-092008).
THE MEMPHIS DATABASE PROJECT
1000–500 BC
35
Visio could be linked with individual datasets and thus form a useful tool for analysing complex genealogical information (see below page 42). After initial research, it became clear that there was currently no commonly available ready-to-use Egyptological database software meeting the requirements as laid down above. Therefore, despite my very limited expertise in the field of software engineering, I finally found myself in the position of having to customise a database environment of my own based on the common Microsoft Access application. EXPERIENCES
During the process of designing the database structure, and also at the later stage of entering the data itself, the advantages as well as the problems connected with a small-to-medium custom-built archaeological database became apparent. This will be illustrated below by a few examples. Maintaining an overview Unquestionably, a general advantage of using databases in Egyptological research lies in the fact that they permit the convenient management of large datasets. The possibility of constant updating is another feature in which a relational electronic database surpasses any kind of printed catalogue. But as has been stated above the benefit obtained through the compilation of detailed and highly structured data on an object or an individual is in danger of being neutralised by the concomitant decrease of transparency. In other words, the user might feel overwhelmed by the great amount of detail and lose the overview. To prevent this the user interface of the Memphis Database provides at the end of the entry form several tables listing key-information about the object (e.g. the divinities or people mentioned and/or depicted) which would otherwise be ‘hidden’ within the detailed accounts on texts and pictorial decoration in the preceding tables (see Figure 2). This information is generated automatically through a dataretrieval query, relieving the user not only of finding the correct data by himself, but also preventing unnecessary data duplication and thereby minimising the loss of data integrity. At the same time, one has to admit that the ‘general commentary’ section, which addresses concerns partly similar to the aforementioned summary tables (i.e. to give a basic idea of the object’s
36
INFORMATION TECHNOLOGY AND EGYPTOLOGY
2008
Figure 2: Memphis Database screenshot showing the summary tables at the end of the entry form
most important features), lies outside the scope of database normalisation.9 This is because data entry here is fairly straightforward and potential updates should not pose any problem to an experienced user. As with the bibliography section, it was decided that a structured set of tables would not be required in this case. The option to conduct full-text searches should be sufficient for retrieving information from there. User considerations and potential online-accessibility As a general rule, the need for a high level of database normalisation and automated decision-making procedures is indirectly proportional to the expertise of the targeted user. Data entered into an amateur database can thus theoretically attain an equal, if not superior, degree of reliability and precision to a professionally-created one if the database is operated by a user well-accustomed both to the topic of research as well as to the software application. On the other hand, this particular fact poses certain problems 9
For the issue of normalisation and general principles of database design, see ADAMS, E.W. And STRUDWICK, N., Relational Database Design: A Tutorial and Case Study for Egyptologists, present volume, esp. pp. 188–9.
THE MEMPHIS DATABASE PROJECT
1000–500 BC
37
as soon as the decision is made to put the database online. Without doubt the only way to use a database to its full potential for dissemination and continuous updating of information is to make it accessible via the internet. However, the increase in the number of potential users (even if only passive ones) necessitates a more professional database management, which in turn depends not only on the developer’s expertise but also on financial backing and consequently cannot be guaranteed for the long term. Data-retrieval and analysis Setting up the database structure and entering the data provides the framework within which the subsequent analytical steps are performed. Fortunately it is relatively easy with Microsoft Access to customise multiparameter queries that can also be implemented into specifically designed search forms. If one were, for example, to investigate potential relations between a Memphite workshop and members of a particular family, one could devise a special query including parameters such as the significant presence or absence of certain logotypes (spelling variants), the different textual components of inscriptions, the material(s) of the object, the occurring personal names of the dedicant and the beneficiary, genealogies, officerelated titles, and so on. In comparison with more traditional methods, such complex queries also greatly enhance the chances of succeeding in disambiguating prosopographical entries of like-named individuals. But as with any other database, practical usage bears out the truism that the quality of the information gained through queries depends upon the quality of the data entered. In the case of the Memphis Database, its quality is largely defined by the quality of the thesauri devised for each data category. This is the reason why great efforts have been undertaken in order to create well-founded typologies and thesaurus hierarchies. Even though I tried to base as many of my thesauri as possible on the already existing Multilingual Egyptological Thesaurus (MET),10 the latter was often found to be too imprecise and unspecific for the purposes of the study. Nevertheless, it 10
The thesaurus was created by Dirk van der Plas and his team and is copyright protected by the Centre for Computer-aided Egyptological Research (CCER). See VAN DER PLAS, D. (ed.), Multilingual Egyptological Thesaurus, Publications interuniversitaires de recherches égyptologiques informatisées 11, Utrecht 1996; an online-version of the MET is available at http://www.ccer.nl/apps/thesaurus/index.html (last accessed on 16-09-2008).
38
INFORMATION TECHNOLOGY AND EGYPTOLOGY
2008
remains an aim to make the newly created thesauri compatible with the MET or to import them into the latter. By doing so one would not only help establishing a generally agreed upon standard of Egyptological artefact classification and nomenclature, but also gain a multilingual interface that could be capitalised on when making the database publicly accessible via the internet. Standardisation vs. faithfulness No artefact, however small, can be represented within a database in its entire complexity. As a consequence, a level of detail has to be chosen that is just high enough to suit one’s requirements. Problems occur when the aim to reduce complexity by standardising features (e.g. for indexing purposes) runs counter to the wish to record a particular cultureme as faithfully as possible and thereby safeguarding against the loss of potentially significant idiosyncrasies. This conflict of interests can be illustrated by the inscriptions on a stela in the Robert Fullerton Art Museum in San Bernardino, CA, belonging to a certain Sheshonq. As Figures. 3a–c show the personal name ‘Sheshonq’ is written in a variety of ways which are not in accordance with the ‘standard’ spelling one would expect (Figure 3a). On the interpretative level of a ‘thick description’, it is perfectly clear that in Figure 3b the personal name ‘Sheshonq’ (S-S-n-q) has to be read, but with a palaeographic agenda in mind the objective would be a faithful recording of what has actually been carved, resulting in the transcription S-S-q-n. However, in doing so another problem occurs. In the common hieroglyphic editors such as Winglyph, the variety of Gardiner N35 in form of a straight line is not available and has to be represented through N17. A similar case is found in Figure 3c. There the last sign resembles a mouth (Gardiner D21) but should rather be taken to represent an egg (H8) for sA (son), indicating a filiation, as can be concluded from the following name of Sheshonq’s father. The question emerges whether a standardising interpretation should be given precedence over recording the phenomenological appearance of a textual element for clarity’s sake. If one were to trace, for example, a particular artisan and therefore were to search for all Memphite funerary stelae on which the filiation sA is written with an egg resembling a mouth, the information on the peculiar shape of H8 has unquestionably have to enter the dataset. The other side of the coin is that greater precision in recording, while enhancing the usefulness of the data for palaeographic analysis,
THE MEMPHIS DATABASE PROJECT
1000–500 BC
39
¥-S-n-q M8A-M8A-n-q-A50 ¥-S-q-n ? M8E-M8E-q:N17-A52A Fig. 3a-c Stela of Sheshonq in the Fullerton Art Museum, San Bernardino, no. 01.001.2002 a - “standard version” of PN Sheshonq
¥-S-(n)-q M8E-M8E-q:r/H8
b - PN in main text, line 4 c - PN in main text, line 6
Figure 3: Stela of Sheshonq in the Fullerton Art Museum, San Bernardino, CA, no. 01.001.2002
unfortunately greatly reduces the comparability of names or other textual elements on the prosopographical level. Though my solution to this problem may neither be very elegant nor meet the highest standards of data normalisation, it is in my opinion the most straightforward one and the easiest to handle. In cases that really matter, textual units should be represented in three different ways: firstly by a photo or an image created with a hieroglyphic editor in order to give an impression of the actual appearance of the word. As the qualitative properties of such an image are not yet susceptible to computerised data retrieval, it is necessary to introduce a second layer, the precise transliteration. This layer accounts for the transfer of the hieroglyphic image into the Manuel de Codage, with faithfully representing potential reversals of sign sequences or misspellings, as well as into the Winglyph code.11 In the third layer, the word is finally represented in a standardised transcription informed by Egyptological interpretation, which makes it usable for general search operations. Taking into consideration that in many cases the second and third layers will 11
Only the latter can be considered a true transcription as it enables to precisely reconstruct the original spelling.
40
INFORMATION TECHNOLOGY AND EGYPTOLOGY
2008
be identical, the entering of data into the third layer should be regarded as compulsory whereas the other layers are merely facultative extensions.
STRUCTURE OF THE DATABASE The major structure of the Memphis Database can be seen in Figure 4. The main form of the user interface is organised into 11 tabs which comprise several sub- and sub-sub-forms. However, not all of them are relevant for each object category. The first two tabs provide information for every kind of object, listing general identifying information, categorisation, measurements, material, date, actual provenance and reconstructed origin, as well as state of preservation. The following tab gives detailed information specific to statues, indicating general type, iconography, attributes, technique of carving and the presence of two-dimensional scenes and texts. In a similar way, tab four is devoted to stelae and offers a detailed form to describe their typological properties. The basic classificatory system for shabtis devised by Schneider12 is implemented in tab five. As with several other data categories of the database, the correct entry of data is facilitated by the possibility to bring on the screen hyperlinked documents providing a graphic overview of the typological features and their respective alphanumeric codes (see Figure 5). The complete string (e.g. 5.3.1 Tc:Cl.XIA2/W38 H4 I8 B 25 Tp3b/V.VIIA) combining the typological sub-codes is generated automatically when entering the specific details. The following two tabs deal with the structure and the rendering of any existing two-dimensional scenes. While tab six offers a basic description of the scenic elements and records their sequence (making the basic layout of a scene susceptible to search operations), tab seven may be used for an in-depth description of any particular scenic element (e.g. a representation of Osiris). It goes without saying that the hierarchical structure of the database with its several sub-forms allows for the fact that a monument may contain many scenes, each of which may comprise numerous pictorial elements worthy of a separate detailed description/classification. Texts are recorded in tab eight, which is itself divided into four subtabs. Among the features recorded are the position of a text on the 12
Cf. H.D. SCHNEIDER, Shabtis. An Introduction to the History of Ancient Egyptian Funerary Statuettes with a Catalogue of the Collection of Shabtis in the National Museum of Antiquities at Leiden, Part I, Leiden 1977.
THE MEMPHIS DATABASE PROJECT
1000–500 BC
41
Structure of database
Identification, object category, measurements, material
Dating, provenance (contextualisation of information is imperative!)
{
Statue: typological criteria Stela: typological criteria
}
Shabti: typological criteria
Scene-description, sequence of pictorial elements, technique
(
Iconography (detail): typological and stylistic description of pictorial element
)
Texts: number, typology, position, sequence of text elements, characteristic logotypes, palaeography, technique
Typological description of pictorial element Text element – detail: epigraphic classification, spelling variants, typology, important information
Summary 1: mentioned/depicted individuals, deities, toponyms etc. (contextualisation)
Summary 2: social context and evaluation of craftsmanship, workshop, bibliographies, comments
Summary 3: text commentary, potential genealogical information
Links to supplementary materials (additional images, documents, etc.)
Figure 4: Basic structure of the Memphis Database
Figure 5: Memphis Database screenshot showing external window helping in the process of data entry
42
INFORMATION TECHNOLOGY AND EGYPTOLOGY
2008
monument, its state of preservation/completeness, a continuous transcription (standardised), significant logotypes (spelling variants of common words and phrases that offer themselves for comparisons), palaeographic property of selected signs, a sequential list of textual elements in order to make the basic composition of a text searchable and comparable to similar texts, techniques of carving, and so on. The potential multiplicity of texts or of textual elements, logotypes, words or signs within a single text is taken care of in a fashion similar to the handling of pictorial elements. Finally, the last three tabs contain summary information automatically retrieved from the scene- and text-sections, a list of primary and secondary publications as well as a large text-box for any general comments on the monument and its features. These last sections are designed to accommodate unstructured free-text entries. The reasons for this decision are as follows: firstly, the bibliography of any given object is not the subject of comparative research and therefore does not require a structured table of its own, and secondly, the full-text searchability of these text-boxes is sufficient for finding a particular publication, an author or a key-word within a commentary.
‘ADD-INS’ In order better to understand the often very complex (sometimes even confusing and contradictory) genealogical data contained in texts an attempt has been made to visualise them by translating them into flowcharts created with Microsoft Visio (see Figure 6). As Visio allows linking any given shape within the chart with an Access dataset, it is possible not only to keep the charts permanently in accordance with the database entries but also to model certain properties of Visio shapes on the latter (e.g. retrieving name and principal titles of an individual from Access data). Whereas the shapes of the Visio flowchart refer to individuals in a manner similar to that of a common genealogical tree, the totality of objects outlined in a particular colour represent the information provided by a single monument (colours are represented in greyscale in Figure 6). Through adding additional layers in different colours the entire network of information associated with a certain family group can be depicted within the chart. By superimposing or masking certain layers, full transparency as to what information is provided by which object is ensured. A critical evaluation of the ancient accounts is thus facilitated. A timeline of royal reigns added to the right of the chart offers the possibility to fix the position of monuments containing regnal
THE MEMPHIS DATABASE PROJECT
1000–500 BC
43
Psammetichus I, year 12
Wnn-nfr People index
People index
8
Title
Hrj pDw.t
Status
mAa-xrw
Psammetichus I
Apr-al-Sdw
Hm.t=f
9
Title
nb.t pr Status
mAa.t-xrw
Psammetichus I sA
Pf-TAw.a.wj-m-BAst.t PA-Ra-ms-sw People index
¤[email protected]@rw §s-As.t-pr.t
Hm.t=f Hm.t=f
Title
nb-t-pr
Status Status
mAa-xrw
sA sA
54
2
Title
jmj-rA aXnwtj
Status
Regnal years
People index
1
mAa.t-xrw mAa.t-xrw
mw.t
Necho II
Nfr-Htp
Regnal years
¡rw-wDA People Status index Title
15 5 mAa-xrw jtj-nTr
Status
anx D.t
Regnal years 6
sA
Psammetichus II Regnal years 19
Pf-TAw.a.wj-m-BAst.t People index Title
Hm.t=f
People index
jmj-rA aXnwtj, jmj-rA pr wr n dwA.tnTr
Status
mAa.t-xrw
mw.t
¡rw-wDA
Amasis
Nfr-rnp.t 5 jtj-nTr
Status
nb-t-pr
Status
sA
Title
2
Title
mAa-xrw
People index
Apries
§s-As.t-pr.t 1
anx D.t
sn smsw
People index Title Status
4
Amasis, year 36
jmj-rA Snwtj anx D.t
Regnal years 2 generations minimum
44 Regnal years 0,5
Figure 6: Genealogical flowchart created with Microsoft Visio
Psammetichus III
44
INFORMATION TECHNOLOGY AND EGYPTOLOGY
2008
years within the absolute (or relative) historical chronology. Thereby the generational intervals of the entire genealogical framework can be adjusted in accordance with the chronological anchors. CONCLUSION AND OUTLOOK
As could hopefully be demonstrated, the Memphis Database Project is not a tool designed to serve exclusively the needs of a PhD dissertation on Late Period Memphis, but has also certain potential to facilitate other Egyptological research in line with the principles of the individual-centred monument analysis. Before the database itself or its core structure can be utilised by other researchers, however, much remains to be done. One of the next steps will consist of implementing a fully integrated extension or counterpart to the existing database in which key information on individuals is grouped together in a dossier style (see also above page 32). It is also clear that it would require the hard work and dedication of people with considerably more IT expertise than I, as well as sufficient funding, to turn this product of an amateur into an efficient and user-friendly tool for data collection and analysis. From an Egyptological point of view, the key questions for the future revolve around the creation, management and extension of the thesauri that are at the very core of the database project and largely determine its quality and relevance. The relation of the already existing thesauri and those yet to come to the MET is an issue of great importance which needs to be clarified before more work is done on these levels. Should archaeological missions working in the Memphite region be willing to contribute their material to the database as well, it would be even more vital to agree upon common typological criteria and terminologies.
EDUCATIONAL IMAGES ON THE WEB
Edward Loring
ABSTRACT
This paper considers the need for making Egyptological photographs accessible for scholars and the general public on the web. Today, practically all relevant institutions have accumulated a considerable number of digital photos, more than they will ever be able to publish in print. It is a great waste of effort if this material is not shared with others through publication on the web. Specific cases of research projects demonstrating the vital importance of images are detailed here and information on web publishing is provided for those who do not have their own web-sites. Sharing images is a win-win situation for all concerned.
The members of I&E stand at the critical and ever more traveled junction between information technology and the humanities. All have, in one way of another, an educational mission. The Internet provides a unique opening for educators to present their material, be it textual or graphic, in a manner which makes it available to all interested individuals and institutions worldwide. It is said that one picture is worth a thousand words. Without a picture many things are difficult or impossible to communicate. Obviously this is especially true in the area of art and to some extent also in the areas of science and technology. A picture combined with text, pointing out a certain element is often the only way to present a convincing argument in professional discussion. Let us take a specific example: Subsequent to the re-excavation of the Royal Cache, TT320, the Centre for Egyptological Studies of the Russian Academy of Sciences (CESRAS) and the Russian Institute of Egyptology in Cairo (RIEC) have pursued a continuing study of the art and artifacts of the 21a Theban 45
46
INFORMATION TECHNOLOGY AND EGYPTOLOGY
2008
Dynasty.1 As the painted coffins of that period are a principal source of cultural information, emphasis has been placed on an analysis of these. In the literature they are generally referred to as ‘yellow coffins’. An in-depth study of digital images made in the Egyptian National Museum in Cairo and in museums of the former Soviet Union reveals that the decoration of many of these coffins, actually the better ones, was painted on a white background and that the present yellow colour was caused by degradation of the pistacia resin varnish applied to protect the surfaces. The generally prevailing dark green colour is a result of this yellowing over designs painted in a light to medium blue. There is considerable resistance among Egyptologists to the white background theory. A number of the later coffins from Bab el-Gasus distributed in European museums do have backgrounds of an ochre wash (not yellow) below yellowed varnish. Technically speaking, the outline draughtsmen would not have been able to draw with water-soluble ink on a varnished surface. The new Armenian Egyptology Centre of Yerevan State University is conducting a project with CESRAS participation to exactly reproduce the painting and varnishing techniques of the period. It would be impossible to prove the case for white backgrounds without detailed colour images. An individual attempting to present the case without such would only be able to refer to specific surface areas of coffins in various museums. The argument for white backgrounds would remain unproven. The necessary images could, of course, be published in printed media but the cost for doing so would be out of the question for any Egyptological journal. In any case, publication in any professional journal with a generally very limited number of copies printed would not make the material available to all Egyptologists, much less to art historians and to a broader culturally oriented public. CESRAS is continually posting evidence for the above argument on its web-site,2 thus making up-to-date research results available to anyone anywhere. Everyone everywhere is able to use our images for their own (noncommercial) purposes. To participate in the project, and is a position to comment on our arguments. We believe that it is the duty of all Egyptological and related institutions to make their visual material available to the public in this manner. 1 2
The term ‘21a Theban Dynasty, ‘D21a’ or ‘21a’ refers to the Theban line of rulers contemporary with those of the 21st Dynasty in Tanis. http://www.cesras.org, under ‘Technology’.
EDUCATIONAL IMAGES ON THE WEB
47
There are many institutions and private individuals with considerable numbers of images which would be of value to Egyptologists in general. Certainly not all of these images are of ‘publication’ quality, but in most cases would none the less be of value to researchers. CESRAS has posted many such images from the early days of digital photography. Admittedly many of them are of very poor quality but they may be the only photos of many objects which will ever be published. A number of the 21a Dynasty painted coffins in museums of the former USSR are in such a miserable state, after seventy years of neglect during the Soviet Period, improper storage, and abuse from transport, that they can neither be moved from their largely inaccessible places of storage, not could they be restored. The problem of moving coffins for complete investigation is another serious problem which we shall address on cesras.org As the volume of information on the web grows, there is a growing necessity for methods to locate specific information efficiently. There are two principle ways to publish images of the web: on web-sites and in photography archives such as Yahoo Flickr. CESRAS is continually posting images in both manners. The basic use of Flickr is gratis. One need only have a gratis Yahoo address to make a private archive. For a small yearly charge a professional membership allows unlimited space for archiving images. The CESRAS Flickr site is http://www.flickr.com/photos/ horemachet. Over three thousand images are to be found there, all in the public domain for free use and the number is growing. Further, CESRAS has a Flickr photo-group with over 4000 images named Egyptology: http:/ /www.flickr.com/groups/egyptology. Flickr members can become members of this group and can then post their images there in a secure archive. A number of excellent photographers participate in the group. All photos are indexed with keywords and can be easily searched. For those who maintain web-sites the best way to make your images searchable is through a considered choice of meta-tags. These are picked up by the various web-crawlers and can be used in combination. The terms used should be short and clear for everybody to understand. The combinations should be well thought out. As dating is vital for Egyptological material, a universal code should be implemented: dynasties tagged D1, D2, D3… An equivalent code should be devised for the reigns of the kings. This is a matter which could be discussed and worked out in an Egyptological forum. The EEF3 would certainly be best for this purpose as many I&E members are regular contributors.
48
INFORMATION TECHNOLOGY AND EGYPTOLOGY
2008
A further possibility to index images or groups thereof is the new ‘open source’ database project at http://www.myowndb.com. This project allows anyone to create their own database at no cost. Each database is a complex matrix which can be searched from any point, much like our GALEXYS system which no one has yet been able to put on the web. We suggest that all those interested in relational databases look into this. It is being managed by dedicated information scientists, and, as it is open source, we can expect good future development. The present search is incremental but we are in communication with the developers and hope to be able to implement the more powerful multidimensional GALEXYS matrix there. The web can and should be used as a tool in research projects requiring photographic information from many sources which could not all be visited by a single researcher. CESRAS has just embarked on a test of this promising method. Here is the specific case. In the study of the coffins mentioned above it is essential to have details on as many objects as possible. The Bab el-Gasus coffins are the most important source of iconographic information. These coffins are scattered about in many locations, some, such as Voronezh, Kazan and Odesssa virtually inaccessible. CESRAS has published, or is publishing, these coffins and has been able to photograph in several important locations. Still, the research material available for our study is very limited. The present study is iconographic with the goal of determining which coffins were decorated by the same artist(s). The study is parametric and will form a structured database of iconographic elements. The first parameter selected is women’s faces and busts. These are being presented on cesras.org and in Flickr for colleagues with access to 21a Theban coffins to study and compare with equivalent images on their examples. Hopefully said colleagues will make photos available on the web to be integrated into the database. This material will then be available to all scholars, students of the graphic arts etc. As time allows, further parameters will be introduced. The next planned is ‘men’s heads’. There it will be interesting to note, for example, that a certain well represented draughtsman (the ‘Vierstrichzeichner’) used completely different lines for male and female faces. This short paper only scratches the surface of the vast possibilities for sharing graphic knowledge on the web. There is more than enough material for all of us and everyone will gain through increased sharing, a win-win situation. 3
http://showcase.netins.net/web/ankh/eefmain.html
DAS GEFLÜGELTE KROKODIL: CODIERUNG VON TOTENBUCH-VIGNETTEN
Marcus Müller-Roth
ABSTRACT
The Book of the Dead is one of the most important and most frequent attested religious texts of Ancient Egypt. Although approx. 50% of all handwritings also show vignettes, which illustrate the spells, the research concentrates nearly exclusively on the texts. Even new editions often confine themselves to a description of the vignettes or offer just an image. A comparative analysis of the iconography or the style is missing. Often parallels are taken into account only in the philological comment. Especially little attention is given to the vignettes of the Book of the Dead of the Saite Recension. In comparison with its precursors of the New Kingdom they are regarded as standardized. Therefore the researchers trust in P. Turin 1791 as a reference. This papyrus was published by Richard Lepsius in 1842 and is regarded as a parallel for the texts as well since that time. But the status of P. Turin 1791 is unclear. Only a small amount of the 1400 known handwritings with vignettes is published. Therefore the multiplicity of the variants is largely unknown. An overview of the material shows that one cannot orientate oneself to P. Turin 1791 without problems. The variants are much more numerous than previously assumed. It can be assumed that local styles are associated with this variants. The article shows how the mass of 1400 manuscripts can be processed to be able to investigate the local variants.
49
50
INFORMATION TECHNOLOGY AND EGYPTOLOGY
2008
1. FORSCHUNGSGESCHICHTE Das Totenbuch zählt zu den bedeutendsten und am häufigsten belegten religiösen Texten des Alten Ägypten.1 Obwohl die Hälfte aller Handschriften neben den Sprüchen auch Vignetten besitzt, die einen erheblichen inhaltlichen und formalen Anteil der Handschriften ausmachen, konzentriert sich die Forschung fast ausschließlich auf die Texte. So werden die Vignetten selbst in aktuellen Editionen, wenn überhaupt, meist nur beschrieben.2 Dagegen vermisst man, dass Stil und Ikonografie wie die Texte anhand von Parallelen analysiert werden.3 Besonders wenig Aufmerksamkeit wird den Vignetten der spätzeitlichen Totenbücher geschenkt, weil sie im Vergleich zu ihren Vorläufern als standardisiert gelten. Von den spätzeitlichen Handschriften mit Vignetten ist jedoch nur ein Bruchteil publiziert, so dass das Spektrum der Varianten weitgehend unbekannt ist. Auch die vor circa 20 Jahren entstandenen Überblickswerke zu den Vignetten leisten nur einen eingeschränkten Dienst. Henk Mildes Dissertation zum Totenbuch des Neferrenpet bietet zwar einen guten Einblick in die Entwicklung vieler Vignetten. Er behandelt auf der Grundlage des ihm 1
2
3
Vgl. T.G. ALLEN, The Book of the Dead or Going Forth by Day. Ideas of the Ancient Egyptians Concerning the Hereafter as Expressed in their own Terms, SAOC 37, Chicago 1974; R.O. FAULKNER, The Ancient Egyptian Book of the Dead (hg. von C. ANDREWS), New York 1972 sowie E. HORNUNG, Das Totenbuch der Ägypter, Die Bibliothek der Alten Welt, Zürich/München 1979 (München 19932). Die vollständige Bibliografie bei S.A. GÜLDEN/I. MUNRO, Bibliographie zum Altägyptischen Totenbuch, SAT 1, Wiesbaden 1998. Eine zweite, erweiterte Auflage ist als SAT 13 in Vorbereitung. Vgl. G. LAPP, The Papyrus of Nu, Catalogue of the Books of the Dead in the British Museum I, London 1997, 58–60; ders., The Papyrus of Nebseni, Catalogue of the Books of the Dead in the British Museum III, London 2004, 53– 5 sowie I. MUNRO, Das Totenbuch des Pa-en-nesti-taui aus der Regierungszeit des Amenemope, HAT 7, Wiesbaden 2001, 60–8. Vgl. I. MUNRO, Der Totenbuch-Papyrus des Hor aus der frühen Ptolemäerzeit, HAT 9, Wiesbaden 2006, 58–73; M. MOSHER, The Papyrus of Hor, Catalogue of the Books of the Dead in the British Museum II, London 2001, 12–22 und 96– 108 sowie M. VON FALCK, Das Totenbuch der Qeqa aus der Ptolemäerzeit (pBerlin P. 3003), HAT 8, Wiesbaden 2006, 59–65. Gegen die Bewertung im letzten Beitrag vgl. M. MÜLLER-ROTH, Lokalkolorit in Schwarz-Weiß, in: B. BACKES/M. MÜLLER-ROTH/S. STÖHR (Hgg.), Ausgestattet mit den Schriften des Thot. Festschrift für Irmtraut Munro zu ihrem 65. Geburtstag, SAT 14, Wiesbaden 2009, 119–31.
DAS GEFLÜGELTE KROKODIL
51
vorliegenden Papyrus aber nur etwa ein Drittel aller Vignetten und konzentriert sich auf die so genannte Thebanische Rezension. Für die Saitische Rezension bietet er dagegen kaum Material.4 Die ausführlichere Arbeit zu den Vignetten der Saitischen Rezension von Malcolm Mosher schafft nur teilweise Abhilfe. Seine Auswertung ist zwar vollständiger und sein Ansatz durchaus lobenswert, aber auf einer Basis von circa 40 Handschriften nur bedingt referenzfähig. Inzwischen sind aus diesem Zeitraum immerhin 1400 Handschriften mit Vignetten registriert. Hinzu kommt, dass die Publikation dieser Arbeit noch immer aussteht und deshalb bisher nur schwer zugänglich ist.5 Bisher liegen auch nur wenige Studien zu einzelnen Vignetten vor. Immerhin widmen sich verschiedene Forscher den großen Vignetten, die die gesamte Höhe des Schriftspiegels einnehmen: Bereits vor 30 Jahren untersuchte Christine Beinlich-Seeber die Szenen des Totengerichts (Tb 125).6 Vor zehn Jahren folgte Judith Gesellensetter mit einer Studie zur Darstellung des Binsengefildes (Tb 110).7 Noch aktueller sind die Arbeiten von Jana Budek zu V 15 8 und von Tarek Tawfik zu V 1.9 Trotz aufkommenden Interesses an den Vignetten der Totenbücher und deren Wertschätzung für die Erforschung der Handschriften herrscht weitgehende Unkenntnis über deren Bandbreite. Deshalb wird aus Mangel an Überblicksstudien blind auf Referenzen wie P. Turin 1791 vertraut, den 4 5
6 7
8
9
H. MILDE, The Vignettes in the Book of the Dead of Neferrenpet, EU 7, Leiden 1991. M. MOSHER, The Ancient Egyptian Book of the Dead in the Late Period: a Study of Revisions Evident in Evolving Vignettes and Possible Chronological and Geographical Implications for Differing Versions of Vignettes, Unpubl. Dissertation Berkeley 1989. Das einzige mir bekannte Exemplar ist auf Mikrofilm in der Universitätsbibliothek Heidelberg einsehbar (Signatur: 2006 RA 3). CHR. SEEBER, Untersuchungen zur Darstellung des Totengerichts im Alten Ägypten, MÄS 35, München/Berlin 1976. J.S. GESELLENSETTER, Das Sechet-Iaru. Untersuchungen zur Vignette des Kapitels 110 im Ägyptischen Totenbuch, Würzburg 1997. URL: http://www.opus-bayern.de/uni-wuerzburg/volltexte/2002/375/. J. BUDEK, Die Sonnenlaufszene. Untersuchungen zur Vignette 15 des Altägyptischen Totenbuches während der Spät- und Ptolemäerzeit, in: SAK 37, 2008, i. Dr. T. TAWFIK, Die Vignette zu Totenbuch-Kapitel 1 und vergleichbare Darstellungen in Gräbern, Unpubl. Dissertation Bonn 2008.
52
INFORMATION TECHNOLOGY AND EGYPTOLOGY
2008
Abb. 1: V 137 in P. Turin 1791
Abb. 2: V 137 in P. Paris BN 129–136
Abb. 3: V 149a in P. Turin 1791
Abb. 4: V 149a in P. Wien ÄS 3862
Richard Lepsius vor über 160 Jahren edierte, und der auch als Referenzgrundlage für Textverweise dient.10 Der folgende Beitrag widmet sich mehreren Fragen: Zunächst wird überprüft, ob P. Turin 1791 seiner Bedeutung als Referenz gerecht wird. Darauf aufbauend wird gezeigt, wie man Vignetten unabhängig vom umgebenden Text identifiziert, wenn dieser fehlt, zerstört ist oder es offensichtlich ist, dass er nicht zur Vignette gehört. In diesem Zusammenhang wird aufgezeigt, dass sich in den unterschiedlichen künstlerischen Ausführungen lokale Stile spiegeln, anhand derer man die Herkunft einer Handschrift ermitteln kann. Abschließend soll vorgestellt werden, wie die Menge von 1400 Quellen aufgearbeitet und bewältigt werden kann, um 10
R. LEPSIUS, Das Todtenbuch der Ägypter, Leipzig 1842 (Neudruck Osnabrück 1969). Vgl. B. BACKES, Wortindex zum späten Totenbuch (pTurin 1791), SAT 9, Wiesbaden 2005.
DAS GEFLÜGELTE KROKODIL
53
Untersuchungen nach obigen Fragestellungen zu ermöglichen und um eine Klassifizierung vorzunehmen.
2. P. TURIN 1791 – EIN STANDARD? Wenn man sich auf einige wenige Handschriften verlässt und von diesen abhängig macht, müssen mehrere Kriterien sichergestellt sein: 1. Die Dokumentation bzw. Publikation der Quelle ist zuverlässig. 2. Zu allen Sprüchen, die grundsätzlich eine Vignette besitzen, existiert ein Beleg.11 3. Die Referenz zeigt den am häufigsten verwendeten Typ der Vignette oder seine Charakteristika. 2.1 Die Zuverlässigkeit der Dokumentation Im Gegensatz zu modernen Editionen, die das Material auf Fototafeln vollständig dokumentieren, zeichnete Lepsius bei P. Turin 1791 sowohl Vignetten als auch Text um. Dabei schlichen sich allerdings Fehler ein. In V 137 trägt der Verstorbene ein spitz zulaufendes Objekt, dass einem Messer ähnelt (Abb. 1). Es handelt sich aber um ein stumpf endendes Sechem-Zepter, wie wir es auch von anderen Quellen kennen (Abb. 2). Ein Blick in die jüngere Fotodokumentation von Boris de Rachewiltz zeigt, dass P. Turin 1791 das Zepter ebenfalls ausführt. Die Umzeichnung von Lepsius ist lediglich unsauber.12 Eine weitere Ungenauigkeit liegt in V 149a vor. Der Dämon des ersten Hügels besitzt in P. Turin 1791 nach Lepsius einen runden Kopf, von dem gerade, kurze Haarstoppel abstehen, und eine lange, spitze Nase, die ihm ein vogelartiges Aussehen verleiht (Abb. 3). Tatsächlich bildet eine Art Pflanzenstaude seinen Kopf, eine tulpenähnliche Mitte umrandet von zwei Uräen (Abb. 4).13
11 12 13
Zu Tb 2–14 und 96–97 existiert prinzipiell keine Vignette. Siehe B. DE RACHEWILTZ, Il Libro dei Morti degli Antichi Egizi. Papiro di Torino, Rom 1986, Tafeln (CXXXVII). Siehe DE RACHEWILTZ, Libro dei Morti, Tafeln (CXLIX). Das Motiv spielt wohl auf Tb 149a, 1 an. Vgl. MOSHER, Vignettes, 413.
54
INFORMATION TECHNOLOGY AND EGYPTOLOGY
Abb. 5: V 19 in P. Paris Louvre E. 7716
Abb. 6: V 42 in P. Leiden T 16
2008
DAS GEFLÜGELTE KROKODIL
55
2.2 Die Vollständigkeit des Papyrus P. Turin 1791 ist unvollständig und weist an zwei Stellen eindeutige Lücken auf. Dabei handelt es sich nicht um Sprüche, die keine spezifische Vignette besitzen14 oder denen in anderen Handschriften eine einfache Darstellung der Verstorbenen als Verlegenheitslösung zugewiesen wurde.15 Vielmehr sind es sehr charakteristische Motive, die wir aus anderen Quellen kennen. Die erste Lücke tut sich bei Tb 19 und 20 auf.16 Dagegen kennen wir aus etwa 60 Quellen zusammen rund 70 Belege für die Vignette der beiden Sprüche.17 Leitmotiv ist der Kranz der Rechtfertigung, der dem Verstorbenen verliehen wird. Am häufigsten steht er dabei dem Gott Atum gegenüber, der dem Verstorbenen laut Tb 19, 1 den Kranz flechtet und überreicht (Abb. 5). Die zweite Lücke besteht bei Tb 42.18 Etwa ein Dutzend anderer Quellen zeigt hier eine Götterreihe aus bis zu 22 Mitgliedern. Sie stehen im Kontext der Gliedervergottung, die Tb 42 behandelt (Abb. 6).19 2.3 Charakteristika in P. Turin 1791 Als Drittes stellt sich die Frage, ob P. Turin 1791 in den einzelnen Vignetten überhaupt die typische bzw. häufigste Form des Motivs überliefert. Eine Überprüfung ergibt, dass der Papyrus über zehn Mal nicht die Ausführung zeigt, die von der Mehrheit der Quellen bevorzugt wird. Manchmal zeigt
14 15 16 17
18 19
Siehe Anm. 11. Siehe Abb. 9 sowie LEPSIUS, Todtenbuch, Tf. XX (V 46), XXI (V 48, 49, 51), XXV (V 65, 66, 67), XXII (V 73) und XXVII (V 76). Siehe LEPSIUS, Todtenbuch, Tf. XIII-XIV. Vgl. E. HASLAUER, Eine Mumienmaske mit dem „Kranz der Rechtfertigung“, in: Jahrbuch des Kunsthistorischen Museums Wien 6/7, Mainz 2006, 233–39; M. MOSHER, Five Versions of Spell 19 from the Late Period Book of the Dead, in: ST.E. THOMPSON/P. DER MANUELIAN (Hgg.), Egypt and Beyond. Essays presented to Leonard H. Lesko, Providence 2008, 237–60 sowie M. MÜLLERROTH, Der Kranz der Rechtfertigung, in: A. MANISALI/B. ROTHÖHLER (Hgg.), Festschrift Assmann zum 70. Geburtstag, i. Dr. Siehe LEPSIUS, Todtenbuch, Tf. XIX. Vgl. S. STÖHR, Who Is Who? Die Repräsentanten der Gliedervergottung in der späten Vignette zu Tb 42, in: B. BACKES/M. MÜLLER-ROTH/S. STÖHR (Hgg.), Ausgestattet mit den Schriften des Thot. Festschrift für Irmtraut Munro zu ihrem 65. Geburtstag, SAT 14, Wiesbaden 2009, 175–200.
56
INFORMATION TECHNOLOGY AND EGYPTOLOGY
2008
Abb. 7: V 43 in P. Turin 1791
Abb. 8: V 43 in P. Paris Louvre N. 3248
Abb. 9: V 51 in P. Turin 1791
Abb. 10: V 51 in P. Paris Louvre N. 3248
P. Turin 1791 nur eine Variante, manchmal ist das gewählte Motiv jedoch stark verändert. Auch hierfür mögen zwei Beispiele genügen: Zu Tb 43 zeigt P. Turin 1791 den Verstorbenen vor drei Göttern, die Was-Zepter tragen (Abb. 7). Diesen Typ für V 43 besitzt immerhin etwa ein Dutzend Quellen. Etwa drei Mal häufiger ist dagegen das gleiche Motiv mit nur einem einzigen Gott. Außerdem fügen fast doppelt so viele Handschriften zwischen dem Verstorbenen und dem Gott drei Köpfe ein (Abb. 8). Sie gehen auf den Text zurück, einem „Spruch, zu verhindern, dass der Kopf des NN ihm abgeschnitten wird im Totenreich.“ Auch zu V 51 liefert P. Turin 1791 ein vereinfachtes Motiv. Er bildet nur die Grundform mit dem Verstorbenen ohne jegliche andere Gestalten oder Attribute ab (Abb. 9). Darin folgen ihm zwar etwa zehn Handschriften und einige weitere geben dem Verstorbenen zusätzlich lediglich einen Stab in die Hand. Acht Quellen erweitern das Motiv dagegen um drei auf dem Kopf stehende Verdammte vor dem Verstorbenen, entsprechend dem Inhalt des Spruchs, „nicht kopfüber zu gehen im Totenreich“ (Abb. 10).
DAS GEFLÜGELTE KROKODIL
Abb. 11: P. Kairo J.E. 97249 (Papyrus 5), Fragmente 1–3
57
58
INFORMATION TECHNOLOGY AND EGYPTOLOGY
2008
3. VARIANTEN Es ist deutlich geworden, dass P. Turin 1791 bei der Identifikation von Vignetten alleine nur bedingt wertvoll ist und selbst ein Überblick über das bisher publizierte Material oft nicht ausreicht. Im Folgenden wird an einigen Beispielen gezeigt, wie nützlich aber auch problematisch es ist, eine möglichst breite Basis an Vergleichsmaterial zu besitzen. Dazu werden zwei Fälle besprochen, in denen zum Vignettenrest kein Text erhalten ist, so dass die Identifizierung allein anhand der Charakteristika der Abbildung erfolgen muss. Beide Fragmente stammen aus Theben-West, wo sie zwischen 1963 und 1969 aus dem Umfeld der Gräber TT 386, 389 und 410 geborgen wurden.20 3.1. Identifizierung der Vignette Das erste Fragment zeigt in der Mitte einen Vogelschwanz, dessen oberes Ende in einen aufrechten Nacken übergeht. Unterhalb des Schwanzes ist noch eine Linie vorhanden, die zum Körper und speziell wahrscheinlich zu den Beinen der Gestalt gehört (Abb. 11, Fragm. 3). Günther Burkard ergänzte die Reste zu einem Ba-Vogel aus V 85. Für diese Deutung sprechen der steile Rückenverlauf sowie die senkrechten Beine, die mit den Vogelgestalten in V 83, 84 und 86 nicht übereinstimmen (Abb. 12).21 Der aufrechte Nacken und der winzige Rest auf der gegenüber liegenden Seite lässt dagegen auf eine Ergänzung nach V 88 schließen. Dieser liegt eigentlich überhaupt keine Vogelgestalt zugrunde, sondern zeigt ein mumifiziertes aufrecht stehendes Krokodil. Die meisten Handschriften führen es ohne weitere Zusätze aus (Abb. 13a). Lediglich eine Minderheit gibt dem Krokodil ein Was-Zepter in die Hände (Abb. 13b). In vielen Handschriften besitzt das Krokodil an seiner Rückseite aber noch einen Fortsatz. Dieser 20
21
Die Gräber publiziert bei D. ARNOLD, Das Grab des Jnj-jtj.f, Grabung im Asasif 1963–1970 I, AV 4, Mainz 1971; J. ASSMANN, Das Grab des Basa (Nr. 389) in der thebanischen Nekropole, Grabung im Asasif 1963–1970 II, AV 6, Mainz 1973 sowie DERS., Das Grab der Mutirdis, Grabung im Asasif 1963–1970 VI, AV 13, Mainz 1977. Zu den folgenden Ausführungen vgl. M. MÜLLER-ROTH, Papyrusfunde aus dem Asasif: Nachträge, in: MDAIK 65, 2009, i. Dr. Vgl. G. BURKARD, Die Papyrusfunde, Grabung im Asasif 1963–1970 III, AV 22, Mainz 1986, 36f. mit Tf. 25c.
DAS GEFLÜGELTE KROKODIL
a) V 83
b) V 84
c) V 85
d) V 86
59
Abb. 12: Vignetten mit Vogeldarstellungen in P. Turin 1791
hat in der Regel die Form eines Krokodilkörpers (Abb. 13c). P. London BM 10315 und P. Lyon H 1579–1583 belegen aber auch eine Variante mit einem Vogelschwanz (Abb. 13d).22 Da es nur zwei weitere Belege für dieses außergewöhnliche Motiv gibt, ist offensichtlich, dass die Vignette auf P. Kairo J.E. 97249 (Papyrus 5) ohne die Kenntnis möglichst aller Belege nicht zu identifizieren ist. Schätzt man den Wert der Identifikation dieser Vignette allein noch gering ein, so liegt aber nahe, dass sich dahinter ein lokales Merkmal verbirgt. Da die vorliegenden Fragmente des P. Kairo J.E. 97249 (Papyrus 5) in situ gefunden wurden, kann der Befund dazu dienen, die beiden anderen Handschriften ebenfalls Theben zuzuweisen. Deren Provenienz ist nämlich unbekannt. Während das erste Beispiel zeigt, wie eindeutig eine Vignette bestimmt werden kann, deren Reste bisher fraglos auf ein anderes Motiv hindeuteten, so kann die Bandbreite des Vergleichsmaterials gleichwohl dazu führen, 22
Der Vollständigkeit halber muss erwähnt werden, dass P. London BM 10097 und P. London BM 10253 das Krokodil in V 88 ausnahmsweise als unmumifiziertes Tier in natürlicher Gestalt wiedergeben. Beide Szenen sind unpubliziert.
60
INFORMATION TECHNOLOGY AND EGYPTOLOGY
a) P. Turin 1791
b) P. London BM 10558
c) P. Berlin P. 3149
d) P. London BM 10315
2008
Abb. 13: Die Vignette zu Tb 88
Abb. 14: pKairo J.E. 97249 (Papyrus 17)
Abb. 15: V 89 in P. Turin 1791
DAS GEFLÜGELTE KROKODIL
61
anscheinend eindeutige Identifikationen zu relativieren, weil das vorliegende Motiv von mehreren Vignetten bekannt ist. Auf einem anderen Fragment ist ein Vogel mit ausgebreiteten Schwingen zu sehen, dessen Kopf jedoch zerstört ist. Darunter sind Reste einer anthropomorphen Mumie erhalten (Abb. 14). Es handelt sich um die prominente Szene, in der der Ba-Vogel über dem aufgebahrten Leichnam schwebt.23 Das Motiv ist zumindest in P. Turin 1791 sowohl aus V 17 als auch V 89 bekannt (Abb. 15). Obwohl von über 200 spätzeitlichen und ptolemäischen Quellen, auf denen Reste von V 17 erhalten sind, nur P. Paris Louvre N. 3081 einen Ba-Vogel über dem Leichnam zeigt, kann eine Entscheidung zugunsten V 89 nicht eindeutig ausfallen.24 Darüber hinaus kommt die Szene in Ausnahmefällen nämlich auch in zwei weiteren Vignetten vor. So zeigen die Achmimer Handschriften P. Hildesheim 5248 und P. MacGregor auch in V 151 einen Ba-Vogel über dem Leichnam, obwohl er dort in der Regel nicht erscheint.25 P. Genf 23464/1–6, P. Kairo J.E. 32887 (S.R. IV 930) und P. London BM 9902 führen den Ba zudem in V 154 aus.26 Obwohl die meisten Belege außerhalb von V 89 Ausnahmen sind, und die Szene einzig für V 89 der Standardtyp ist, so ist doch klar, dass das Motiv des Ba-Vogels über dem Leichnam prinzipiell über allen Szenen erscheinen kann, in denen der aufgebahrte Leichnam vorkommt. Natürlich kann die Provenienz, sofern sie bekannt ist, die Wahrscheinlichkeit einer Identifikation erhöhen. Trotzdem ist eine sichere Zuordnung grundsätzlich fraglich. Die Beispiele zeigen, dass eine Übersicht über die Bandbreite der Variationen nicht nur eine Neuidentifikation zur Folge haben kann, sondern scheinbar eindeutige Identifikationen relativieren kann. Um hierüber einen schnellen Überblick zu gewinnen und allgemein für die Arbeit mit Vignetten, regte mein Kollege Burkhard Backes an, analog zum Wortindex, mit dessen Hilfe man Totenbuchpassagen selbst winziger Fragmente identifizieren kann, einen entsprechenden Motivindex zu erstellen.27 Dort würde man unter dem Stichwort „Ba-Vogel“ alle Vignetten finden, in denen der Ba-Vogel in jedweder Variante Bestandteil der Vignette ist. 23 24 25
26 27
Vgl. BURKARD, Papyrusfunde, 68–71 mit Tf. 66,8. Siehe P. BARGUET, Le livre des Morts des anciens égyptiens, LAPO 1, Paris 1967, 59. Siehe B. LÜSCHER, Das Totenbuch pBerlin P. 10477 aus Achmim (mit Photographien des verwandten pHildesheim 5248), HAT 6, Wiesbaden 2000, Photo-Tafel 36 sowie MOSHER, Hor, Tf. 16.1. Alle drei Szenen unpubliziert. Vgl. BACKES, Wortindex.
62
INFORMATION TECHNOLOGY AND EGYPTOLOGY
2008
3.2 Identifizierung der Provenienz Wie oben bereits angedeutet, ist es nahe liegend zu vermuten, dass die Merkmale der Varianten ihren Ursprung in lokalen Stilen haben. Neben chronologischen Unterschieden ist die Provenienz oder die Werkstatt sicherlich ursächlich für die Varianten. Prinzipiell steigt die Wahrscheinlichkeit, dass es sich bei einem Merkmal um eine lokale Eigenheit handelt, gewiss mit zunehmender Individualität. Je individueller das Merkmal und je deutlicher die Unterschiede, desto stärker ist außerdem das Unterscheidungskriterium und damit die Aussagekraft des Befundes. Die Vignette zu Tb 149 ist hierfür ein aussagekräftiges Beispiel: Der Dämon der ersten Höhle erscheint in sehr unterschiedlichen, sehr individuellen Ausführungen. Die Verteilung der Quellen zeigt deutlich, dass es sich um lokale Typen handelt: In Theben verwendet man die bereits oben kennen gelernte Ausführung mit einer Pflanzenstaude auf dem Kopf (Abb. 16a). Daneben existiert in einer Minderzahl ein anthropomorpher Kopf, aus dem zwei federähnliche Objekte ragen (Abb. 16b). In Memphis ist der Kopf des Dämons völlig anders gestaltet. In der Regel besitzt der Dämon dort einen Bes-Kopf (Abb. 16c).28 Eine einzige Ausnahme zeigt eine Darstellung, die wohl einen abgeschlagenen Kopf und Blutfontänen zeigen sollen (Abb. 16d). Die mittelägyptischen Handschriften heben sich von beiden geografischen Polen ab. In Herakleopolis nähert man sich dem thebanischen Typ mit rundem Kopf an, führt ihn jedoch schwarz aus (Abb. 16e). Die oberen Fortsätze sehen wie abstehende Haarstoppel oder Grashalme aus. Die Quellen aus Achmim reduzieren die Gestalt auf eine kahlköpfige anthropomorphe Gestalt (Abb. 16f). Ein dritter Typ ersetzt den Kopf durch Messer, die im Hals stecken (Abb. 16g). Das Beispiel zeigt, dass Handschriften der gleichen Herkunft einen gemeinsamen Vignettentyp besitzen. Während sich hier alle Handschriften auf sehr unterschiedliche Typen verteilen lassen, handelt es sich in der Regel um Varianten und Merkmale, die sich in kleiner Anzahl von der Menge absetzen. So zeigt Fragment 1 des oben bereits behandelten P. Kairo J.E. 97249 (Papyrus 5) einen Vogelschwanz, über dem ein Schattenwedel schwebt (Abb. 11, Fragm. 1). Bei diesem Schattenwedel handelt es sich um eine seltene aber charakteristische Variante von V 26. Dort sitzt ein Ba-Vogel auf 28
Bei MOSHER, Vignettes, 413 mit Tf. 210 gar nicht erwähnt, obwohl der abgebildete P. Paris Louvre N. 3091 die Darstellung zeigt.
DAS GEFLÜGELTE KROKODIL
a) P. Wien ÄS 3862
b) P. Ryerson
c) P. Portheim (A)
d) P. Louvre N. 3081
e) P. Colon. Aeg. 10207
f) P. Berlin P. 10478
g) P. Milbank Abb. 16: Übersicht über die Typen von V 149a aus Theben (a-b), Memphis (c-d) und Mittelägypten (e-g)
63
64
INFORMATION TECHNOLOGY AND EGYPTOLOGY
2008
Abb. 17: V 26 in P. Turin 1791
Abb. 18: Die Eigenschaften der Vignetten zu Tb 19 und 20 (P. Paris Louvre E. 7716)
einem Sockel, während ein hockender Adorant ihm gegenüber sein Herz in der Hand hält (Abb. 17). Den Fächer führen nur P. Berlin P. 3008, P. Boston MFA 92.2582, P. Dublin 1663, P. Kairo J.E. 97249 (Papyrus 5), P. London BM 9976, P. Lyon H 1579–1583 und P. Paris Louvre E. 7716 aus. Wertet man die Informationen dieser Handschriften bezüglich ihrer vermeintlichen Herkunft aus, erkennt man, dass es sich beim Schattenwedel um eine thebanische Variante handelt.
4. AUFARBEITUNG Sowohl die kritische Betrachtung des P. Turin 1791 als auch die Darlegung des Variantenreichtums haben gezeigt, dass das publizierte Primärmaterial und die bisher vorliegende Sekundärliteratur immer nur einen Teileinblick bietet. Gerade bei seltenen Varianten wie dem geflügelten Krokodil aus
DAS GEFLÜGELTE KROKODIL
65
V 88 (Abb. 13d) ist diese Grundlage unzureichend, um Identifikationen sicher vorzunehmen. Immerhin handelt es sich allein für den Zeitraum der Saitischen Rezension um etwa 8000 einzelne Vignetten. Zählt man bei Vignetten, die aus mehreren Bildern oder Szenen bestehen, diese Teile jeweils einzeln, summiert sich die Menge sogar auf ca. 15000 Abbildungen. Ein Überblick ist momentan allein im Totenbucharchiv in Bonn möglich.29 Wie kann das Material so zugänglich gemacht werden, dass einerseits ein Überblick über diese Masse möglich ist und zum anderen die Varianten hinsichtlich ihrer Provenienz, ihres Alters oder anderer Kriterien ausgewertet werden können? Dabei ist zu beachten, dass eine einfache Unterteilung anhand nur eines Kriteriums, wie Mosher dies betrieb, meist unzureichend ist.30 Viele Bestandteile der Vignetten können nämlich variieren und lokale Eigenheiten widerspiegeln. Die Varianten zu V 19 und 20 veranschaulichen das Problem (Abb. 18).31 a) Als Konstellation kann der Verstorbene dem Gott Atum gegenüber stehen. Er kann aber auch einer anderen Person gegenüber stehen oder alleine auftreten. b) Der Kranz der Rechtfertigung kann als Leinenstreifen ausgeführt sein oder als geflochtener Reif. Manche Vignetten ersetzen ihn sogar durch einen Halskragen, wie er auch in Varianten von V 158 bekannt ist.32 Einige Quellen verzichten auch völlig auf den Kranz. c) Das Podest, auf dem der Kranz liegt, ist zum einen als einfaches Rechteck ausgeführt, zum anderen als Torbau gestaltet. Manchmal handelt sich auch um einen Tisch. Außerdem kann das Podest fehlen, wenn der Kranz von einer Person dargereicht wird. 29
30 31 32
URL: http://www.totenbuch-projekt.uni-bonn.de. Zur Geschichte des Projekts H. KOCKELMANN, From One to Ten: The Book of the Dead Project after its First Decade, in: B. BACKES/I. MUNRO/S. STÖHR (Hgg.), TotenbuchForschungen. Gesammelte Beiträge des 2. Internationalen Totenbuch-Symposiums 2005, Bonn, 25. bis 29. September 2005, SAT 11, Wiesbaden 2006, 161–65. Siehe Anm. 5. Siehe Anm. 17 mit der relevanten Literatur. So P. Kairo J.E. 32887 (S.R. IV 930), P. Langres, P. Leiden T 18, P. London BM 10983, P. Paris Louvre N. 3084, P. Paris Louvre N. 5450, P. St. Gallen, P. Wien ÄS 3862 + 10159, P. Wien Vindob. Aeg. 10.110, M. Berlin o. Nr. (PsammetichMeri-Neith), M. Princeton, Pharaonic Rolls, No. 8, M. Uppsala o. Nr. (Nofretiu), M. Sydney R 397. Siehe URL: http://libweb2.princeton.edu/rbsc2/ papyri/Bookofthe DeadRoll8.html (M. Princeton, Pharaonic Rolls, No. 8).
66
INFORMATION TECHNOLOGY AND EGYPTOLOGY
2008
Abb. 19: Die Eingabemaske für die Eigenschaften der einzelnen Vignetten
DAS GEFLÜGELTE KROKODIL
67
Abb. 20: Mumienbinde Prag, Náprstek Museum K 249
d) Zwischen den Akteuren kann ein Opfertisch stehen oder fehlen. e) Die Personen können preisend dargestellt sein oder etwas darreichen. Sie können auch einen Stab in der Hand halten. Unabhängig von ihrer Gestik sitzt der Verstorbene in Ausnahmefällen auch auf einem Stuhl. Hinter jedem dieser Merkmale kann sich eine lokale Eigenheit verbergen. Nach welchem muss man nun die Typen einteilen, um relevante Befunde zu erhalten? Je nach Klassifizierung könnten dabei völlig unterschiedliche Ergebnisse entstehen. Muss man anhand aller Merkmale eine Einteilung vornehmen? Wie kann dies bei einer Menge von 1400 Quellen geschehen? Die Lösung besteht in der Festlegung der einzelnen Merkmale und ihrer statistischen Auswertung. Zunächst muss man die Eigenschaften festlegen: Welche Personenkonstellation liegt in der jeweiligen Quelle vor? Wie ist der Kranz dargestellt? Welche Form hat das Podest? Ist ein Opfertisch vorhanden? Welche Gesten besitzen die Gestalten? Diese Eigenschaften müssen in einer Datenbank registriert werden. Dazu wurde die im Totenbuch-Projekt bereits existierende FilemakerDatenbank erweitert. In einer neuen Tabelle werden nun die Eigenschaften aller Vignetten separat registriert. In einer weiteren Maske werden den einzelnen Vignetten diese Eigenschaften zugewiesen (Abb. 19). Sind die Merkmale der Vignetten derart verzeichnet, kann man die Vignetten und ihre zugehörigen Quellen nach allen verwendeten Kriterien sortieren. Die Quellen der vorliegenden Vignette zu Tb 19 und 20 kann man beispielsweise zunächst anhand der Personenkonstellation trennen. In einem zweiten Schritt kann man die Quellen zudem anhand der Darstellung des Kranzes unterteilen.33 Diese Verteilung ergibt für die Darstellung des Verstorbenen ohne Atum (a) und der Interpretation des Kranzes als Leinenstreifen (b) eine kleine Gruppe aus fünf Handschriften (Tab. 1). Vier
68
INFORMATION TECHNOLOGY AND EGYPTOLOGY
2008
der fünf Quellen werden bereits Memphis zugeordnet. Quellen aus anderen Regionen fehlen dagegen. Deshalb ist die Wahrscheinlichkeit groß, dass M. Prag K 249 (Abb. 20), dessen Herkunft bisher unbekannt war, ebenfalls aus diesem Raum stammt. P. Kairo CG 40029 P. Paris Louvre N. 5450 P. Wien ÄS 3862 + 10159 P. Wien Vindob. Aeg. 10.110 M. Prag, Náprstek Museum K 249
Memphis Memphis Memphis Memphis unbekannt
V 20 V 20 V 19 V 19 V 19
Tabelle 1: Ergebnis einer Abfrage nach V 19 und 20, die den Verstorbenen ohne Atum zeigen und den Kranz als Leinenstreifen abbilden P. Kairo CG 40029 (J.E. 95837, S.R. IV 934) P. Kairo CG 40029 (J.E. 95837, S.R. IV 934)
Memphis Memphis
V 19 V 20
P. Wien ÄS 3862 + 10159 P. Wien ÄS 3862 + 10159 P. Wien Vindob. Aeg. 10.110 M. Florenz Inv. 3681 M. Prag, Náprstek Museum K 249
Memphis Memphis Memphis Memphis unbekannt
V 19 V 20 V 19 V 19 V 19
Tabelle 2: Ergebnis einer Abfrage nach V 19 und 20, die das Podest, auf dem der Kranz der Rechtfertigung liegt, als Tisch darstellen
Eine andere Verteilung kann anhand der Form des Podests (c) vorgenommen werden. Das Ergebnis zeigt, dass ein kistenartiges Podest nur in Theben verwendet wird. Auch die Form als Torbau ist fast nur aus Theben bekannt. Die Quellen, bei denen der Kranz auf einem Tisch liegt, stammen dagegen fast alle aus Memphis (Tab. 2). Hierzu gesellt sich auch wieder M. Prag K 249. Somit gibt es allein durch V 19/20 zwei Befunde, die für M. Prag K 249 eine memphitische Herkunft plausibel machen. Bestätigt wird das Ergebnis durch das Layout der Mumienbinde, die Holger Kockelmann als Formular 1b bezeichnet.34 Hierunter versteht er Mumienbinden mit im Text eingebetteten Vignetten, die eine Rahmung 33
Das Folgende ausführlicher bei M. MÜLLER-ROTH, Der Kranz der Rechtfertigung, in: A. MANISALI/B. ROTHÖHLER (Hgg.), Festschrift Assmann zum 70. Geburtstag, i. Dr.
DAS GEFLÜGELTE KROKODIL
69
besitzen. Die Bindenhöhe beträgt bei dieser Gruppe 5–8 cm, auf denen etwa 5–7 Zeilen untergebracht sind. Die trifft auch auf M. Prag K 249 zu, die 6 cm hoch ist und 5 Zeilen trägt (Abb. 20). Laut Kockelmanns Untersuchung, der das gesamte Corpus der Mumienbinden zugrunde liegt, stammt dieses Layout aus Memphis und Gurob.
5. FAZIT Die oben festgestellten Befunde verdeutlichen, dass die Vignetten wichtige Hinweise zur Herkunftsbestimmung der Totenbücher liefern können. So dienen die lokalen Vignettentraditionen und die damit verbundenen ikonografischen und stilistischen Varianten dazu, die Provenienz der bisher nicht zugewiesenen Handschriften zu erschließen. Andere Hinweise liefern unter anderem die Titeln der Totenbuch-Besitzer oder das Layout.35 Das oben gezeigte Beispiel zeigt, dass M. Prag K 249 aus Memphis stammt, da er das Podest als Tisch ausführt sowie den Verstorbenen alleine zeigt und mit dem Kranz der Rechtfertigung in Form eines Leinenstreifens kombiniert (Abb. 20). Natürlich ist der Befund einer einzigen Vignette nur bedingt aussagekräftig. Wenn aber nur eine einzige Vignette erhalten ist, wie es bei M. Prag K 249 der Fall ist,36 gewinnt man durch die Ikonografie der Vignette einen wertvollen Hinweis. Ist mehr Material erhalten, sollten natürlich mehrere Befunde ermittelt werden, um die Argumentation zu festigen.37 Je größer die Anzahl an Vignetten und Merkmalen, die aussagekräftige Ergebnisse liefern, desto eindeutiger ist natürlich der Befund. 34
35 36
37
Vgl. H. KOCKELMANN, Untersuchungen zu den späten Totenbuch-Handschriften auf Mumienbinden, Band II: Handbuch zu den Mumienbinden und Leinenamuletten, SAT 12, Wiesbaden 2008, 95f. Vgl. M. MOSHER, Theban and Memphitic Book of the Dead Traditions in the Late Period, in: JARCE 29, 1992, 145–51. Vgl. M. VERNER, Verejné sbírky staroegyptskych ´ památek v CSSR, Vol. II, Unpublizierte Magister-Arbeit Prag 1964. Laut KOCKELMANN, Untersuchungen, 275, Anm. 273 gehört M. Paris, Louvre E. 18861 zum gleichen Set. Ob darauf auch Vignetten erhalten sind, ist weder ihm noch mir bekannt. Vgl. M. MÜLLER-ROTH, Papyrusfunde aus dem Asasif: Nachträge, in: MDAIK 65, 2009, i. Dr. sowie DERS., Lokalkolorit in Schwarz-Weiß, in: B. BACKES/M. MÜLLER-ROTH/S. STÖHR (Hgg.), Ausgestattet mit den Schriften des Thot. Festschrift für Irmtraut Munro zu ihrem 65. Geburtstag, SAT 14, Wiesbaden 2009, 119–31.
70
INFORMATION TECHNOLOGY AND EGYPTOLOGY
2008
Da von den 1400 Quellen mit Vignetten aus der Saitischen Rezension nur von etwa 550 die Herkunft bekannt ist, ist das Potential dieser Analyse offensichtlich. Selbst die manuelle Verteilung anhand nur eines oder zweier Kriterien ermöglicht es mir schon heute etwa 100 Quellen neu zuweisen. Es ist damit zu rechnen, dass sich diese Zahl durch eine computergestützte Auswertung, wie sie oben vorgestellt wurde, verdoppeln lässt. Das gezeigte Vorgehen ermöglicht außerdem, auch sehr komplexe Vignetten wie die Darstellung des Totengerichts zu analysieren und zu klassifizieren. Vignetten mit mehreren Szenen sind manuell nämlich nur schwer zu untersuchen. Neben der Vielzahl an Szenen, die jeweils unterschiedliche Varianten besitzen können, kann zusätzlich auch die Anordnung und Abfolge der verschiedenen Szenen variieren.38 ABBILDUNGSNACHWEIS
Abb. 1: LEPSIUS, Todtenbuch, Tf. LVI; Abb. 2: Zeichnung M. Müller-Roth; Abb. 3: LEPSIUS, Todtenbuch, Tf. LXXI; Abb. 4: Zeichnung M. Müller-Roth; Abb. 5: Zeichnung M. Müller-Roth; Abb. 6: C. LEEMANS, Papyrus égyptien funéraire hiéroglyphique (T. 16) du Musée d’Antiquités des Pays-Bas à Leide, Monumens égyptiens du Musée d’Antiquités des Pays-Bas à Leide III, 4, Leiden 1876, Tf. 15; Abb. 7: LEPSIUS, Todtenbuch, Tf. XX; Abb. 8: BARGUET, Livre des Morts, 86; Abb. 9: LEPSIUS, Todtenbuch, Tf. XXI; Abb. 10: BARGUET, Livre des Morts, 89; Abb. 11: Zeichnung M. Müller-Roth; Abb. 12: LEPSIUS, Todtenbuch, Tf. XXXI–XXXII; Abb. 13a: LEPSIUS, Todtenbuch, Tf. XXXIII; Abb. 13b—d: Zeichnung M. Müller-Roth; Abb. 14: Zeichnung M. Müller-Roth; Abb. 15: LEPSIUS, Todtenbuch, Tf. XXXIII; Abb. 16: Zeichnung M. Müller-Roth; Abb. 17: LEPSIUS, Todtenbuch, Tf. XV; Abb. 18: Zeichnung M. Müller-Roth; Abb. 19: Screenshot; Abb. 20: Zeichnung M. MüllerRoth.
38
Vgl. Anm. 6–9.
AUTOMATIC ALIGNMENT OF HIEROGLYPHS AND TRANSLITERATION
Mark-Jan Nederhof
ABSTRACT Automatic alignment has important applications in philology, facilitating study of texts on the basis of electronic resources produced by different scholars. A simple technique is presented to realise such alignment for Ancient Egyptian hieroglyphic texts and transliteration. Preliminary experiments with the technique are reported, and plans for future work are discussed.
1. INTRODUCTION A convenient form to represent analysis of a manuscript is as interlinear text. In this form, the text is divided into fragments, each short enough to fit within the width of a page or of a computer screen. We will refer to such fragments as phrases, which may or may not concur with the linguistic meaning of the term. For each phrase, a number of rows present different aspects of the phrase, which may be the original text, some form of transcription, word-by-word gloss, translation, or a combination of these types of data. The data that occupies the i-th row of the interlinear text for each phrase is called a tier, or sometimes stream. In the case of Ancient Egyptian, interlinear text typically offers three tiers, consisting of hieroglyphs, transliteration and translation. Additional tiers may offer glosses and lexical or syntactic analyses. The hieroglyphic text may be a facsimile, but more often, we find a normalised transcription using an electronic font, especially when the original manuscript is in 71
72
INFORMATION TECHNOLOGY AND EGYPTOLOGY
2008
hieratic. The text direction is often mirrored with respect to the original manuscript to be left-to-right, to match the directionality of the other tiers. By transliteration we mean the rendering of the text using the modern Egyptological alphabet, which is composed of some letters from the Latin alphabet in combination with diacritics, and two additional letters representing aleph and ayin. Interlinear text commonly offers only one translation in one modern language, but considering that many interpretations of some of the more difficult texts are still contentious, it can be very fruitful to compare several different translations, displayed as consecutive tiers. This may also be said about transliteration, especially where the segmentation of a hieroglyphic text into words is uncertain. In general, one particular interpretation of a hieroglyphic text is best represented by the combination of transliteration and translation. Many applications of interlinear text involve audio recordings. Such a recording in an appropriate visualisation can be one of the tiers, but it may also serve as the basis for annotations. For example, occurrences of words in a transcription as well as prosodic units can be mapped to time intervals within the recording. Alignment of such annotations can be done straightforwardly through the total ordering imposed by the time line of the recording. Several annotations can be compiled by different linguists, allowing automatic creation of interlinear text, typically restricted to a selection of the tiers, depending on the interests of the user. A survey of tools and techniques involving such applications was presented by Bird and Liberman (2001). They pointed out that annotations can also be mapped to offsets within a particular textual resource, in place of anchor points within an audio recording. This requires however that the textual resource is unchanging, and that different scholars agree on the choice of this textual resource. These constraints regrettably preclude use in many branches of philology. In the example of Ancient Egyptian texts, it would be impractical to demand that all scholars who translate or annotate a text should tag their resources with indices in some canonical representation of the text. Note that a hieroglyphic transcription as interpretation of an hieratic text cannot serve as such a canonical representation, because there may not be any such interpretation that has the approval of the entire community. Existence of lacunas would further exacerbate the problem.
AUTOMATIC ALIGNMENT OF HIEROGLYPHS
73
If the creation of interlinear text cannot rely on anchor points in a common resource offering a total ordering, then an obvious alternative is to align different textual resources automatically, by analysing the contents of the tiers. Alignment of, for example, English, French and German translations of Egyptian texts can be done by relatively conventional techniques; see for example Gale and Church (1993). The present paper will focus on automatic alignment of hieroglyphs and transliteration, which is a form of monolingual alignment involving two very different writing systems. The implementation of this task is one significant component within a larger system to create interlinear text out of one or more hieroglyphic transcriptions, transliterations, translations, and lexical and syntactic annotations. In passing, we would like to point out that similar techniques can also be applied to automatic alignment of different manuscripts of the same text. Examples are the four manuscripts of the Eloquent Peasant, the dozens of manuscripts coverings parts of Sinuhe and the countless manuscripts offering different versions of the Book of the Dead. Alignment of different manuscripts of the same text entails specific problems. For example, a phrase in one manuscript may be absent in another, or entirely different phrases may occur in the respective manuscripts. Even more difficult to handle automatically are cases where the same phrases occur, but in a different order. These issues will not be addressed in any detail here. The task of automatic alignment of hieroglyphic text and transliteration is related to the automatic transliteration of hieroglyphs, which was investigated in a seminal paper by Rosmorduc (2001). He used finite-state transducers, achieving very high accuracy. Whereas automatic alignment seems an easier task in comparison, it is still far from trivial, especially as we have decided not to involve lexica or grammatical knowledge. The rationale is that incorporating such knowledge could bias certain genres or periods, and make the software less robust. Another related task is word segmentation, which means dividing a sequence of signs into words. It differs from our alignment task in that the words themselves are not known. Word segmentation is relevant in general for writing systems without explicit word boundaries. It has received much attention for Chinese. Most conventional algorithms for word segmentation rely on the availability of lexica; see e.g. Sproat et al. (1994). Again, this is incompatible with our objectives. The structure of this paper is as follows. Section 2 discusses the ongoing activities that form the context to the work reported here. The
74
INFORMATION TECHNOLOGY AND EGYPTOLOGY
2008
orthographic model underlying the automatic alignment is discussed in Section 3 and initial experiments are discussed in Section 4. Section 5 outlines plans for further work.
2. CONTEXT 2.1 an XML format for alignment The XML format AELalign allows encoding of: • hieroglyphs, • transliteration, • translation, and • lexical annotation. For one manuscript, there may be several hieroglyphic transcriptions, several transliterations, etc., and these may be distributed over different files. Moreover, several manuscripts for the same text may be included. Constraints on alignment can be explicitly indicated by line numbers in the manuscripts, or by additional anchor points relating one tier to another. For more details, see Nederhof (2002a). A first trial of its use involved a joint effort over the World Wide Web to translate the Eloquent Peasant with a group of students. Participants submitted their interpretations of parts of the text by email, in a very simple plain-text format, containing transliterations, translations and comments. This format was automatically converted to AELalign. In a next phase, the given hieroglyphic text, which was also in the AELalign format, was aligned with the respective interpretations to form an interlinear text in HTML, which could be viewed as a web page. This served as a virtual blackboard, allowing joint discussions about different interpretations. After this successful trial, small adjustments were made to the format, and the viewing software was reimplemented to provide output in PDF and in a Java applet. The implementation in Java provides the most flexibility, allowing a selection of the tiers to be displayed. The amount of text that fits on each line depends on the width of the window, and as soon as the window size is changed, suitable line breaks are determined anew, leading to a new interlinear text. An excerpt from the PDF output is given in Figure 1. We see that the hieroglyphic text is conveniently divided into parts that are aligned with phrases consisting of transliteration and translation. Until recently, such precise alignment could only be achieved by manually inserting suitable
75
AUTOMATIC ALIGNMENT OF HIEROGLYPHS
M ¯ z i( ÁVÊ /L z B1 | L¯y 42
42 B1 | PNZMUQ+PR NV;WM 42 B1 | 'Look, I shall take away your donkey, peasant, R
M ¯ z L¯/L z
R PN+PR N R 'But look, your donkey
43
5 vk¸Ë
| 43 +U | ZQP I6PR M 43 because | it ate my barley. 9.6
V
| 9.6 | +U 9.6 | is
Ë
v5 ZQPMW M
eating my barley!
Figure 1: Part of interlinear text showing two versions of the Eloquent Peasant
anchor points. From Section 3 onward, it will be explained how precise alignment can be done automatically.
2.2 HIEROGLYPHIC ENCODING The hieroglyphic encoding we use is called the Revised Encoding Scheme (RES), and represents a significant departure from the Manuel de Codage (MdC) from Buurman et al. (1988). The main shortcomings of MdC encoding of hieroglyphs are: • There is no precisely defined standard independent from any software tool. • The syntax is chaotic and common interpretations of the official documents seem to entail ambiguities. • The operators are not nearly expressive enough to represent a fair portion of the relative positioning of signs one finds on good monumental inscriptions. • The Manual de Codage seems to be the product of feature creep by having it dictate not only the encoding of hieroglyphs themselves but also the layout of a document that contains hieroglyphs, as well as rudimentary grammatical annotations. A few key properties of RES are: • The syntax is very simple, and the meaning is rigorously defined. Given a string of characters, it can be decided with certainty whether it is or is not a valid fragment of hieroglyphic encoding, and if so, its visualisation is fully prescribed, with the font and a small number of other parameters as free variables.
76
INFORMATION TECHNOLOGY AND EGYPTOLOGY
2008
• In place of absolute positioning, introduced by some dialects of MdC to make up for shortcomings in its expressivity, a number of operations are available in RES that allow the composition of signs to be described as one sees them, and a suitable appearance can be automatically computed on the basis of a given font. This means that the validity of an encoding can survive a change of font. • RES basically only involves hieroglyphs and not other types of text. The only exceptions are footnote markers next to hieroglyphs (whose exact positions are determined automatically), and brackets for philological purposes. An example of the enhanced expressive power is: insert[te](G39,N5) ~% The meaning of this use of the ‘insert’ operation is that N5 (‘sun’) is placed in the empty right-upper corner of G39 (‘pintail’). In particular, N5 is scaled down as much as necessary to leave a default distance between the two signs. (This distance can be adjusted if desired. With distance 0, the two signs are touching.) The reason the validity of this construction may survive a change of font is that the positioning and scaling depend on the sizes and shapes of the individual signs. For example, in a font where the right-upper corner of G39 leaves less empty space, the occurrence of N5 would be scaled down more. (It should be pointed out that a similar construction exists in PLOTTEXT, developed by Stief (1985).) This should be contrasted with the corresponding notation in most dialects of MdC, using an ampersand. The above example would be written G39&N5. This construction is called a ‘ligature’ or ‘special group’. Both terms are misleading, because the individual signs are not joined together as in traditional ligatures, and there is nothing special about such groups, considering they are quite common in any hieroglyphic text. The meaning of the so-called ligatures, in terms of the relative positioning and scaling of signs, is fixed in the font or in the software. Either way, no standardisation is achieved by the notation itself, and different tools could assign different meanings to ligatures. Attempts to exhaustively list all ligatures and prescribe standardised meanings are futile, as any newly found long text will very likely contain ligatures not included in any fixed list. A case in point is the EGPZ sign list, which contains no less than 400 ligatures.1 While investigating an MdC encoding of Papyrus Westcar, which is 1
Version 1.0, November 2007, at http://www.egpz.com/resources/egpz.htm.
AUTOMATIC ALIGNMENT OF HIEROGLYPHS
77
one of the most popular Middle Egyptian texts, we found two ligatures that were absent from the EGPZ list. Our comments carry over to superimposition of signs. Instead of requiring a proliferation of combined signs as separate code points as in the case of most MdC dialects, RES offers the ‘stack’ operation, as for example in: stack[on](V28,I9) The introduction of RES by Nederhof (2002b) has not been well received by the Egyptological community. The main objections raised by members of the audience, and by others before and after the meeting, were: 1. The MdC is generally accepted as the standard, and too many existing encoded texts would become obsolete if RES were adopted. 2. The goal of preserving the validity of an encoding across different fonts, which is one of the strengths of RES, is irrelevant because Egyptologists typically throw away an encoding once they have published a text. In other words, the electronic encoding is no more than an intermediate form towards a final product on paper. 3. RES is too verbose. Instead of C2\ as in MdC, one must write C2[mirror]. 4. The uniform syntax of RES is irrelevant, as typical users only approach hieroglyphic encoding via a graphical interface. 5. RES is not an XML format. 6. Even the precise placement of signs relative to each other as allowed by RES would not suffice for palaeographic purposes. 7. The rendering of, for example, the ‘insert’ operation is too expensive and too complicated for some applications. The first objection is in conflict with the second, and at least one of them must be invalid. The same holds for the third objection versus the fourth and the fifth objections. Apart from this, each of the above allows a number of counter-arguments. The first objection can be rejected by pointing out that MdC is not a standard. Various tools exist today that each implement one possible interpretation of part of the features from Buurman et al. (1988), and these interpretations vary widely. All of these tools further extend MdC by new features, to make up for shortcomings in its expressivity. However, as different tools add different such features, encoded texts created with one tool become obsolete as soon as that tool becomes obsolete, and exchanging encodings across different tools is problematic.
78
INFORMATION TECHNOLOGY AND EGYPTOLOGY
2008
To be able to benefit from existing texts encoded in MdC, we have implemented a tool to automatically convert various MdC dialects to RES. As very different principles underlie MdC and RES, respectively, manual post-processing is regrettably required in most cases. As to the second objection, the reason why encodings of hieroglyphs are considered to be ephemeral may be just because the grave inadequacies of available formats such as MdC have so far hindered the development of any large electronic corpora of hieroglyphic texts. Considering the corpora available in other areas of philology, including those involving non-alphabetic writing systems, such as Akkadian and Sumerian, it is unclear why the particular qualities of Ancient Egyptian would preclude the creation of similar corpora in Egyptology, to be freely shared among different scholars. The syntax of RES is more verbose than that of MdC, in the sense of requiring more characters to describe the same thing, but this helps to make the constructions more self-explanatory, and the main objective was to cast the enhanced expressive power into a uniform syntax. The simplicity of the syntax of RES may not be appreciated by end-users as much as by developers of hieroglyph-processing tools, which counters the fourth objection above.2 As to the fifth objection, an XML version of RES will be created as soon as an immediate need for it arises, which has not been the case since RES was introduced. The sixth objection is based on a misunderstanding of what RES wants to achieve. The purpose of an electronic encoding is to offer a visual appearance somewhere between a purely linear sequence of hieroglyphs on the one hand, which would be utterly unacceptable to any scholar, and a facsimile of the original manuscript on the other, which would be impractical in applications involving e.g. interlinear text. RES does not have the pretences to replace facsimiles, but it does move further away from an unacceptably rigid and unrealistic partition of the text surface into perfect squares as MdC would have it. Furthermore, it cannot be denied that developers and users of MdC software in the past have felt a strong need for more accurate scaling and positioning of signs. In fact, after the introduction of RES, some MdC tools have adopted some of its features and added them to their dialects of MdC. 2
One striking observation illustrating the relative complexity of MdC notation is the following. The specification of the tokeniser for MdC in Serge Rosmorduc’s JSesh is 188 lines long, against 34 lines for RES in our Java implementation.
AUTOMATIC ALIGNMENT OF HIEROGLYPHS
79
Regrettably, this exacerbates some of the other problems of MdC, such as lack of standardisation and the chaotic syntax. In response to the seventh objection above, we have introduced RESlite, in which signs receive absolute values for their positioning and scaling. In applications where the font is fixed, RES and RESlite offer the exact same visual appearance, and both allow a fragment of hieroglyphic text to be divided into smaller fragments, e.g. to allow line breaks where hieroglyphs are part of running text. In practice, RES is the format most suitable for exchange between groups of scholars, whereas RESlite can be used internally in systems to allow quick rendering using very simple software, following one-off automatic conversion from RES to RESlite. As far as automatic alignment is concerned, the choice of RES as opposed to MdC for the hieroglyphic encoding is not essential, because the implementation as yet ignores relative positioning of signs beyond a purely linear order. Nevertheless, RES is preferable for this task, due to its emphasis on standardisation and avoidance of ad hoc signs and ligatures. Moreover, RES is ideal for interlinear text, allowing automatic line breaks and padding, and explicitly providing for applications to enforce a horizontal left-to-right text direction irrespective of the encoded directionality.
3. MODEL Experienced Egyptologists would have little difficulty in correctly aligning hieroglyphs with corresponding transliteration. As with any other problem in the realm of artificial intelligence however, it is not so easy to capture expert knowledge in a formal representation allowing the same task to be done reliably by mechanical means. Whereas alignment seems straightforward in the case of idealised input, many problems arise in practice. For example, some occurrences of signs may have a non-standard reading not listed in any grammar or dictionary. Further, there may be errors, made by the modern scholar in the hieroglyphic encoding or in the transliteration, or errors by the ancient scribe not reflected in the transliteration. An alignment algorithm should therefore be designed to avoid a complete failure of the task when confronted with input that is less than ideal. In particular, upon encountering problematic writings, local errors may be unavoidable, but these should not spread to cause incorrect alignments for larger parts of a text.
80
INFORMATION TECHNOLOGY AND EGYPTOLOGY
2008
The ability of software to give reasonable output even when the input suffers from a limited number of inadequacies is known as robustness. In general, the more complicated an algorithm is, the more difficult it is to achieve robustness. We have therefore started our investigations by choosing a very simple ‘orthographic’ model of how hieroglyphic signs are combined to write words, in terms of their transliteration. This model assumes only two classes of signs, namely phonograms and determinatives. Ideograms will be treated as phonograms with the special property that they can only match against the start of a word. For example, sign D56 (‘leg’) will be treated as a phonogram rd that can only match against the first two consonants of a word (or more precisely, of a morpheme; see further below). We will refer to a mapping from signs to collections of possible readings as an ‘annotated sign list’. In some cases, the mapping is from a sequence of signs to one or more readings; for example, three consecutive occurrences of N35 (‘ripple of water’) may together have a reading as phonogram mw. In the experiments, reported in Section 4, we have extracted our annotated sign list from the ‘Zeichenliste’ of Hannig (1995). It is relatively straightforward to map this list to a data structure that is machine readable. As we wanted to make the experiments reproducible and eliminate subjective decisions as much as possible, Hannig’s list seemed preferable to the one from Gardiner (1957), which would have left much more room for interpretation. It should be noted that Hannig’s sign list is less complete than Gardiner’s. In particular, many uncommon readings of signs are absent. This does not hinder our experiments however, and in fact, the existence of gaps in the sign list helps us to measure the robustness of the algorithm, in the light of the awareness that no sign list will ever cover all readings of all occurrences of signs in unseen texts. With a fixed annotated sign list, the actual input to the alignment algorithm consists of a sequence of hieroglyphic signs and a sequence of words in transliteration. The order of the signs is roughly as they occur in the hieroglyphic encoding in RES. An exception is made however for the ‘insert’ operation, where the order depends on whether the inserted sign is placed before or after the main sign. The alignment algorithm to be described below reads the hieroglyphic text from beginning to end, maintaining positions, which represent the boundaries between pairs of consecutive hieroglyphs, plus the position
81
AUTOMATIC ALIGNMENT OF HIEROGLYPHS
w
ntr 0
R8
1
wj
wj
ntr
ntr
R8
2
R8
3
Figure 2: Edges between positions indicate possible readings of signs or sequences of signs. For example, sign R8 (‘cloth wound on a pole’) can be read as phonogram nTr . The second occurrence in sequence can alternatively be read as the dual ending wj, and the second and third occurrences can together be read as the plural ending w. Hence a path from position 0 to position 3 exists with the edges nTr and w, respectively, which can be matched against a word nTrw . (To simplify the figure, other readings, such as the feminine dual and plural endings, were omitted.)
before the first hieroglyph and the position after the last hieroglyph. Positions are connected by edges labelled by the possible meanings of the hieroglyphs between those positions, as determined by the annotated sign list. A meaning is either a string of consonants for a reading as phonogram (or ideogram, as explained before), or it is simply the information that a sign can serve as determinative. Special treatment is needed for numbers and for dual and plural. For a sequence of numerals between two positions, an edge is added between those positions, labelled by the corresponding number in decimal notation, as it might occur in the transliteration. For two consecutive occurrences of the same sign, edges are added labelled by phonograms wj and tj with the extra constraint that they can only match the final two consonants of a word. Something similar holds for plural, in the case of three occurrences of the same sign, as exemplified in Figure 2. The words of the transliteration are simply defined as strings separated by white space, consisting of consonants and punctuation signs (i.e. ‘.’, ‘-’, or ‘=’). No attempt was made to do automatic morphological analysis beyond the explicit punctuation signs. For example, the feminine ending t is treated like any other consonant, as our transliteration conventions, which follow Hannig (1995), do not mark the boundaries between stems and feminine or plural endings.
82
INFORMATION TECHNOLOGY AND EGYPTOLOGY
2008
The morphemes that are separated by punctuation signs are treated as individual entities however where it concerns our model of orthography. The basic assumption is that a morpheme is written as a sequence of phonograms, together covering all consonants in the transliteration, from left to right, followed by a sequence of zero or more determinatives. By ‘left to right’ we mean that the first consonant covered by a phonogram should not follow any consonants that have not yet been covered by previous phonograms. In order to achieve robustness, orthographic analyses are allowed that violate the above basic assumption, at the cost of a ‘penalty’, the height of which depends on the seriousness of the violation, based on our intuitions about hieroglyphic writing. For example, a phonogram which follows rather than precedes a determinative incurs a penalty of 8. If a semi-vowel (j or w) in the transliteration is not covered by any phonogram, this incurs a penalty of 2, while this penalty is 5 for other consonants. A hieroglyphic sign that is ignored altogether incurs a penalty of 20. The task is now to automatically determine how consecutive hieroglyphic signs corresponding to words in the transliteration. This is realised by going through the hieroglyphic signs from beginning to end, jumping from position to position following the edges, while at the same time going through the words from the transliteration from beginning to end. The labels of the edges are matched against words from the transliteration, which may incur penalties as outlined above. One difficulty is however that the correct alignment of hieroglyphs and transliteration is not known in advance, and at each moment, it may be decided to terminate the recognition of the current word of the transliteration and move to the next. Our approach is to pursue all possibilities in parallel, and in the end the solution is returned that minimises the sum of the incurred penalties. More precisely, we define a configuration as a triple consisting of the following three components: 1. A position in the sequence of hieroglyphic signs, as explained before. 2. Precisely one of the following: • A position in the sequence of words. Positions are defined much as in the case of hieroglyphs. Each represents the boundary between a pair of consecutive words, and there is one position before the first word and one position after the last word.
AUTOMATIC ALIGNMENT OF HIEROGLYPHS
83
• An occurrence of a word in the transliteration, together with an indication which of the consonants have been covered by phonograms encountered earlier. 3. The sum of penalties so far. At the beginning of the alignment algorithm, we have one configuration, with penalty 0, pointing to the beginning of the hieroglyphic text and to the beginning of the transliteration. We call this the initial configuration. New configurations are derived from existing ones by different steps. The main steps are: • The recognition of one word is finalised, moving to the position between the current word and the next. • From a position between two words, the recognition of the next word is initiated. • We follow an edge between two hieroglyphs, moving to a next position. In the case of a phonogram, the corresponding consonants in the current word are marked as having been covered. • We ignore an hieroglyphic sign, moving to the next position. We say a configuration is final if it simultaneously points to the end of the hieroglyphic text and to the end of the transliteration. Of all final configurations, the one is taken that has the smallest penalty. By tracing back how the final configuration originated, one indirectly obtains a preferred matching of sequences of hieroglyphs against words in the transliteration. The algorithm applies two tricks that allow the task to be done within a few seconds, even for long texts. First, where two competing configurations are identical except for their penalties, the one with the highest penalty is discarded. This can be easily justified, as the configuration with the higher penalty will certainly not be part of the optimal solution when we reach a final configuration. This trick falls within a range of techniques that are known as ‘dynamic programming’. Secondly, for each position within the hieroglyphic text, we only consider the configurations with the N lowest penalties among all configurations associated with that position. Here N is a low number, for example 40. This technique is known as ‘beam search’. The rationale is that partial solutions that seems less promising than many competing partial solutions will likely not be part of the optimal solution in the end. Although beam search is very effective in truncating useless computations, there is a risk that the optimal solution itself is truncated. To reduce this risk, N should be chosen sufficiently high.
84
INFORMATION TECHNOLOGY AND EGYPTOLOGY
2008
Figure 3 shows an example of some configurations that match three consecutive hieroglyphic signs against an occurrence of word xprt, starting from a configuration that points to position 41 just before the corresponding hieroglyphs and to position 82 just before the word, with an overall penalty of 10, which is the sum of the penalties incurred earlier. From this configuration, another is derived that points to the word xprt between positions 82 and 83, and position 41 as before. This configuration contains information about which consonants have already been covered by phonograms. In the figure this is indicated as a hyphen (‘not yet covered’) or an asterisk (‘covered’). At the beginning we have only hyphens. After L1 is interpreted as phonogram xpr, a new configuration is obtained, pointing to position 42 and to the word xprt between positions 82 and 83 as before, now with three asterisks for the three covered consonants. From here, one may process D21 as phonogram r and X1 as phonogram t, and then finish recognition of the word, leading to the configuration with overall penalty 10 as before, pointing to positions 44 and 83. Alternatively, the recognition of the word may be terminated just before D21 is processed, and then the total penalty increases by 5 for the t in xprt that is not accounted for. The resulting configuration has overall penalty 15, and points to positions 42 and 83. More penalties seem unavoidable after that, as D21 and X1 may need to be skipped in order to process following words in the transliteration, and each skipped hieroglyph carries a penalty of 20. Note that the higher the overall penalty becomes, the more likely it is that the configurations will eventually be discarded in favour of competing configurations with lower penalties. A feature was built in to deal with simple cases of honorific transposition, involving a single sign R8 (‘cloth wound on a pole’), N5 (‘sun’) or M23 (‘swt-plant’) to be moved across one or more words of the transliteration. This is realised by allowing such a sign to be skipped and stored in a ‘buffer’ in a configuration, to be retrieved from a later configuration derived from it. No additional mechanism was needed to deal with transposition for honorific or aesthetic purposes within single words, as the basic orthographic model is already fairly permissive with regard to the order of signs within words (although this by itself causes some errors, as we will see in the next section). Honorific transposition in general may involve a god’s name written with several signs. It is not clear how to deal with this without slowing down
85
AUTOMATIC ALIGNMENT OF HIEROGLYPHS
t hpr ˘ 41
L1
hprt ˘- - - 10
10
81
ktt
jt
r 42
hprt ˘∗∗∗10
82
D21
43
15 hprt ˘∗∗∗10
hprt ˘
X1
44
35 hprt ˘∗∗∗∗ 10
83
m
10
84
Figure 3: The circles at the top represent positions within the hieroglyphic text, those at the bottom represent positions within the transliteration. The rectangles are configurations, each containing the sum of penalties so far and a pointer to a position in the hieroglyphs. Each of the small rectangles also points to a position between two words in the transliteration. The large rectangles each point to an actual word in the transliteration, while indicating which of the consonants have been covered by phonograms so far. The dotted arrows indicate how one configuration is derived from another. By following such arrows backwards, one can find out how the final configuration with the lowest penalty was obtained from the initial configuration, through a list of steps that identifies the preferred alignment between hieroglyphs and transliteration. (Only those configurations are depicted here that are relevant to the discussion in the running text.)
86
INFORMATION TECHNOLOGY AND EGYPTOLOGY
2008
the alignment algorithm considerably, and therefore we have not attempted to solve the general case in the current implementation.
4. EXPERIMENTS The first text that was considered is the Shipwrecked Sailor. It was found to be very suitable for experimentation with different variants of the alignment algorithm, as it may be the least complicated of all the longer Middle Egyptian texts, having only minor lacunas and few problematic readings. The annotated sign list was not changed after we started our experiments, but changes were made to the software at this stage. This means that the error rates cannot be taken as typical for unseen texts of the same level of difficulty, let alone unseen texts of higher levels of difficulty due to, for example, unusual writings of words. We produced a hieroglyphic encoding of the text, and a transliteration that closely follows the conventions of Hannig (1995), the same dictionary from which the annotated sign list was extracted. By these conventions, the text is 1014 words long. A compound word consisting of two parts connected by a hyphen was counted as one word. Also suffix pronouns were not counted separately. In a first phase, we segmented the hieroglyphic encoding manually, marking the first sign of the writing of each word. A simple graphical user interface was developed to help this process, allowing signs to be marked by mouse clicks, while putting the corresponding word from the transliteration under the position of each marked sign, and showing the next few words from the transliteration. In a second phase, the automatic alignment was run to find the first sign corresponding to each word. This was compared to the manual alignment, and the graphical user interface then identified the differences by highlighting. Some auxiliary tools were added to provide explanations why certain mismatches between manual and automatic alignment arose. This includes a tracer, showing the steps of the alignment process for a selected part of the text. Among the 1014 words, only 12 errors were made by the automatic alignment. These can be divided into 8 errors that are due to gaps in the annotated sign list, and only 4 that are due to inadequacies of the orthographic model. Examples of gaps in the sign list are the absence of the reading of A50 (‘man of rank seated on chair’) as ideogram for Spsj, and the
AUTOMATIC ALIGNMENT OF HIEROGLYPHS
87
absence of the reading of A12 (‘soldier with bow and quiver’) as ideogram for mSa. The crudeness of the orthographic model accounts for the failure to match F20 (‘tongue of ox’) with reading as phonogram ns against a subsequence of consonants in the word nj-sw. One error occurs in xprt.n rdjt, where the second occurrence of D21 (‘mouth’) is matched to the r from xprt.n rather than the r from rdjt; the problem here is that the model does not pose enough restrictions on the order of phonograms. In two occurrences of snTr (‘incense’) the honorific transposition of R8 (‘cloth wound on a pole’) misleads the model into taking the sign as determinative of the preceding word. Whereas each of the above errors could clearly be eliminated by an ad hoc patch of the model, it seems likely that every unseen text will raise new problems, and a 100% accuracy is beyond reach. Furthermore, a frequent observation in computational linguistics is that tweaking models to correctly handle specific cases may inadvertently lead to other cases being handled incorrectly. Moreover, increasing coverage, for example, by adding possible readings to the sign list, may well lead to a decrease in accuracy. On the positive side, for each of the cases discussed above, no trailing errors in subsequent words ensued. This means the algorithm is very robust, in the sense that local errors do not tend to spread to larger parts of the text. Moreover, for purposes of producing interlinear representations, it may not be a cause for great concern to have the start of a word misidentified by a distance of only one or two signs. In a second experiment we investigated Papyrus Westcar, repeating the above procedures. This was done after all parameters of the model had been fixed. This means that the results can be seen as typical for unseen texts of the same level of difficulty. However, due to the many lacunas, it was often problematic to identify the sign occurrence where we would want the automatic alignment to find the beginning of a word. Mismatches between manual and automatic alignment that arose as a direct result of lacunas have therefore been ignored, leaving 81 errors, among the 2683 words of the transliteration. Of these errors, 24 are due to gaps in the annotated sign list, and the remaining 57 must be blamed on inadequacies of the orthographic model. Among the latter, the most frequent problem is honorific transposition within a single word, accounting for 33 errors. Of these, 14 occur in the writing of nsw-bjtj and 6 in the writing of snTr.
88
INFORMATION TECHNOLOGY AND EGYPTOLOGY
2008
The other inadequacies of the orthographic model were found at pairs of consecutive words sharing one or more consonants, accounting for 24 errors. Most notably, in 14 occurrences of Dd.jn Ddj the first occurrence of R11 (‘column imitating a bundle of stacks’), with reading as phonogram Dd, is incorrectly taken as part of the writing of the first word Dd.jn, rather than the second word Ddj, which has two occurrences of R11. (See Gardiner (1957, 502) for the reading of two consecutive occurrences of R11.)
5. FUTURE WORK The above reported preliminary results from work in progress. The orthographic model we have described allows a large spectrum of refinements, and the current project plans to pursue several of them. This includes evaluation on the basis of a wider range of texts. A first priority will be the creation of a sign list that contains more detailed and accurate annotations on possible readings of signs. Although the recent Unicode proposal (Everson and Richmond, 2007) greatly contributes to the standardisation of signs used in electronic encoding of hieroglyphs, it is regrettable that no accompanying document is currently being planned that summarises and updates the information about the signs collected by Gardiner (1957; as well as several other documents). It cannot be emphasised enough that electronic resources offering such information are of the highest importance to automatic processing of hieroglyphic texts. The creation of annotated sign lists in an electronic format also forces us to look closer at the different classes of signs and their functions in the writing of words. Whereas some Egyptian grammars distinguish between only three different classes of signs, viz. phonograms, ideograms (also called logograms) and determinatives, some publications use a finer distinction. Schenkel (1971) in addition offers a formal description of how words are composed of signs with various functions. This description cannot be readily employed for our purposes however, as no sign list exists that is annotated with corresponding functions. Furthermore, Schenkel’s work does not directly link hieroglyphic writing to transliteration. For example, it does not specify how to deal with phonetic complements. The annotated sign list that we used in the experiments was derived from the ‘Zeichenliste’ of Hannig (1995). The original list distinguishes between Phon, Log, Abk, Det, Phono-Det, and Log/Det. Whereas the informal meanings of these concepts may be clear, it is less obvious what
AUTOMATIC ALIGNMENT OF HIEROGLYPHS
89
functions these classes of signs should have in a formal model of orthography. With refinements of the orthographic model, the mechanism of penalties described in Section 3 may become harder to maintain. The finer the constraints are that one imposes on the orthography, the more frequently will constraints be violated by valid orthographic analyses of actually occurring hieroglyphic text, and thereby penalties may be incurred in all competing analyses. The selection of the desired analysis will therefore often depend on a suitable choice of the relative heights of different kinds of penalties. Regrettably, human intuitions tend to be quite unreliable when it comes to estimating quantitative aspects of language or, in this case, writing systems. We therefore need to investigate stochastic approaches, to replace penalties by probabilities that are automatically estimated on the basis of annotated or unannotated hieroglyphic texts. Due to the nature of the writing system, which lacks unique standardised spellings, and due to the sparsity of the data, it would be infeasible to estimate the probability of each possible variant spelling of each word separately. A more promising approach is to compute parameters that abstract away from the actual consonants of a word in transliteration, looking at the order in which, for example, phonograms are used to represent the consonants in respective positions. As an example, consider the writing of nst (‘throne’) as: C∑ N35 : F20 - X1 : W11 4® The exact probability of this writing, given the word in transliteration, is: P(N35,F20,X1,W11 | nst) = P(N35 | nst)⋅ P(F20 | nst,N35)⋅ P(X1 | nst,N35,F20)⋅ P(W11 | nst,N35,F20,X1)⋅ P(end | nst,N35,F20,X1,W11). For example, the third factor in the right-hand side of this equation should be read as the probability that X1 is the third sign in the writing of nst, following the signs N35 and F20 in this order. The final factor represents the probability that the word ends after the given list of four signs. Whereas accurate estimation of each of the factors in the above is infeasible, we can approximate them by for example: P(N35,F20,X1,W11 | nst) ≈ P(*-- | ---)⋅⋅
90
INFORMATION TECHNOLOGY AND EGYPTOLOGY
2008
P(**- | *--)⋅⋅ P(--* | **-)⋅⋅ P(det | ***)⋅⋅ P(end | ***), given the information that N35 can be a phonogram n, which concurs with the first consonant of nst, W11 can be a determinative, etc. Each minus sign or asterisk in the above represents a position in the word in transliteration. The asterisks in the left-hand sides of factors of the form P(⋅ | ⋅) are the consonants covered by the next phonogram. The asterisks in the right-hand sides indicate the consonants that have been covered by previous phonograms. For example, P(--* | **-) represents the probability that the next sign is a phonogram matching the third consonant of a three-consonant word, given that the first and second consonants have already been covered by previous phonograms. There are many variants of such a model. For example, probabilities can be conditioned on the previous one or two signs (cf. bigrams and trigrams), or appropriate abstractions from those signs, making use of ‘smoothing’ of probabilities in the case of sparse training data. Very similar techniques exist for other applications in computational linguistics, such as part-of-speech tagging (Manning and Schütze, 1999). So far we have assumed that hieroglyphic text is processed as a linear list of signs, without indication of the exact relative positioning. In particular, line breaks and the separations between quadrats are ignored. There are cases however where relative positioning is essential to the correct reading of hieroglyphs. One classical example is m-Xnw written with N35a (‘three ripples of water’) below W24 (‘bowl’); see Gardiner (1957, 134). In the investigated texts, no examples were found of incorrect alignment of hieroglyphs and transliteration that could be amended if relative positioning beyond a purely linear order were to be taken into account. It cannot be excluded however that relative positioning could help to increase the accuracy of alignment. Signs may generally be grouped together following aesthetic principles, irrespective of how a sequence of signs is to be segmented into words. For example, if the last sign of one word and the first sign of the following word are both roughly one quadrat in width and half a quadrat in height, they may be grouped together into a single quadrat, with one sign above the other. An interesting conjecture by Horst Beinlich (personal communication) is however that there was a certain tendency to let the boundaries
AUTOMATIC ALIGNMENT OF HIEROGLYPHS
91
between consecutive words concur with boundaries between consecutive quadrats. This merits further investigation. To the extent the conjecture may be confirmed by the data, it holds the potential to enhance the accuracy of automatic alignment. We have found that the penalties discussed in Section 3 sometimes signal errors in the hieroglyphic encoding, often due to a confusion between signs with similar appearances. Another suggestion for further research is therefore to develop tools that highlight potential errors in hieroglyphic transcriptions. Lastly, it should be pointed out that the most useful applications of automatic alignment are at this moment hindered by the fact that many hieroglyphic transcriptions and translations are available only in printed form. It is highly desirable, for this reason and for many others, that scholars in the future will make more of their textual resources available in suitable electronic formats, either free of copyright or at least explicitly allowing use within viewing software.
6. CONCLUSIONS Whereas the work reported here is in early stages, some conclusions can already be drawn. First, automatic alignment of hieroglyphs and transliteration is feasible with very simple techniques, without using lexica or grammatical knowledge. The accuracy may vary across texts, but experiments show that at least some texts allow a very high accuracy. In addition, there is ample room for refinements of the discussed techniques, with the potential to further reduce the error rate. Second, our work underlines the importance of standardisation of hieroglyphic encoding. In addition, the creation of electronic resources, such as annotated sign lists documenting the possible functions of signs in the writing of words, is essential for automatic processing of texts.
7. ACKNOWLEDGEMENTS Many of the presented ideas were inspired by unpublished work by Serge Rosmorduc on automatic transliteration, and I am greatly indebted to him for many fruitful discussions. Much gratitude goes to Nigel Strudwick for his technical assistance with the typesetting of this article. I am also very
92
INFORMATION TECHNOLOGY AND EGYPTOLOGY
2008
grateful to Horst Beinlich for discussions on the orthography of Egyptian, and to Norbert Stief for correspondence about PLOTTEXT. This article was written with the generous assistance of a fellowship from the Leverhulme Trust.
REFERENCES J. Buurman, N. Grimal, M. Hainsworth, J. Hallof, and D. van der Plas. Inventaire des signes hiéroglyphiques en vue de leur saisie informatique. Institut de France, Paris, 1988. S. Bird and M. Liberman. A formal framework for linguistic annotation. Speech Communication, 33: 23–60, 2001. M. Everson and B. Richmond. Proposal to encode Egyptian hieroglyphs in the SMP of the UCS. Working Group Document ISO/IEC JTC1/ SC2/WG2 N3237, International Organization for Standardization, 2007. A.H. Gardiner. Egyptian Grammar. Griffith Institute, Ashmolean Museum, Oxford, 1957. W.A. Gale and K.W. Church. A program for aligning sentences in bilingual corpora. Computational Linguistics, 19(1): 75–102, 1993. R. Hannig. Grosses Handwörterbuch Ägyptisch-Deutsch: die Sprache der Pharaonen (2800–950 v.Chr.). Verlag Philipp von Zabern, Mainz, 1995. C.D. Manning and H. Schütze. Foundations of Statistical Natural Language Processing. MIT Press, 1999. M.-J. Nederhof. Alignment of resources on Egyptian texts based on XML. In Proceedings of the 14th Table Ronde Informatique et Egyptologie, 2002a. On CD-ROM. M.-J. Nederhof. A revised encoding scheme for hieroglyphic. In Proceedings of the 14th Table Ronde Informatique et Egyptologie, 2002b. On CD-ROM. S. Rosmorduc. Transducteurs pour la translittération des hiéroglyphes. Unpublished paper presented at TALN 2001, 2001. W. Schenkel. Zur Struktur der Hieroglyphenschrift. Mitteilungen des deutschen archäologischen Instituts, Abteilung Kairo, 27: 85–98, 1971. R. Sproat, C. Shih, W. Gale, and N. Chang. A stochastic finite-state wordsegmentation algorithm for Chinese. In 32nd Annual Meeting of the Association for Computational Linguistics, Proceedings of the Conference, pages 66– 73, Las Cruces, New Mexico, USA, 1994. N. Stief. Hieroglyphen, Koptisch, Umschrift, u.a. – ein Textausgabesystem. Göttinger Miszellen, 86: 37–44, 1985.
CORPUS ÉLECTRONIQUES DE L’ANCIEN ÉGYPTIEN: TRAITEMENT XML DES TEXTES DES PROCESSIONS DE SOUBASSEMENT DES TEMPLES TARDIFS
Vincent Razanajao
ABSTRACT
An important element of the grammar of the temple, the processions of Nile Gods inscribed on the lower parts of walls in temples of Ptolemaic and Roman date form an interesting corpus for the study of toponyms or for the history of religions. When the author came to consider the form the publication of this corpus should take, the idea of an electronic publication and treatment by XML seemed very promising. The aim of this paper is not to elaborate the Document Type Definition (DTD) dedicated to Egyptian texts but to put forward the author’s comments subsequent to having used the corpus. The subject of the project being textual corpora, the approach will follow that of the Text Encoding Initiative (TEI). After a brief history of processions and an analysis of their structure, a series of adjustments of the TEI tags will be presented with an explanation of the bias chosen. The outline of an XML file is appended at the end of the paper.
Les réflexions d’ordre informatique dont je souhaite faire part dans les quelques pages qui vont suivre sont nées des questionnements apparus lors de l’établissement d’un projet d’édition des textes des processions ornant les soubassements des temples tardifs, principalement d’époques grecque et romaine. Lorsqu’il s’est agi de réfléchir à la forme que devait prendre ce corpus, l’idée d’une publication électronique a tout de suite paru opportune et,
93
94
INFORMATION TECHNOLOGY AND EGYPTOLOGY
2008
plus précisément, un traitement par XML de ces textes a semblé tout à fait intéressant. L’aspect électronique répond à la volonté d’aller au-delà de la publication des textes pour elle-même, l’objectif étant de développer un outil permettant une interrogation multiple et illimitée de la matière contenu dans ces textes. La possibilité serait ainsi offerte à tout chercheur d’obtenir aisément, par exemple, l’ensemble des textes relatifs à la province d’Égypte sur laquelle il travaille. LE XML, SES PRINCIPES ET SES APPLICATIONS DANS LES ETUDES DE CORPUS TEXTUELS
Principes généraux du XML Le XML, ou «eXtensible Markup Language», a pour principe d’augmenter toute chaîne de caractères – et donc tout texte – de «balises» invisibles permettant d’ajouter des informations sur les mots qui la composent. Par exemple, le balisage XML de la notice «égyptologie» du Petit Robert consisterait à indiquer que: -le premier mot, «égyptologie», est l’entrée de la notice; -«mil. XIXe» est la date de la première occurrence; -«de Égypte et –logie» constitue l’étymologie; -«étude scientifique de l’Égypte ancienne» en donne le sens. Les balises XML étant de petites séquences entre signes < et >, une traduction en XML de cet exemple donnerait: Égyptologie, mil. XIXe, de Égypte et –logie. Étude scientifique de l’Égypte ancienne. Les règles de saisie XML Le traitement d’un texte selon les normes XML passe par la définition préalable du canevas et des règles qui seront utilisés pour la saisie. L’ensemble de ces règles est répertorié dans ce qui s’appelle une DTD, ou «Document Type Definition», une sorte de document modèle qui énonce les règles propres à la mise en XML des textes du corpus correspondant.
CORPUS ÉLECTRONIQUES DE L’ANCIEN ÉGYPTIEN
95
XML et corpus textuels: les préceptes de la TEI Traiter de manière informatique le corpus qui nous intéresse nécessite de s’inscrire dans les problématiques développées par la TEI, ou Text Encoding Initiative (http://www.tei-c.org/), «que l’on pourrait traduire, ainsi que cela a été très justement écrit1, par groupe d’initiative pour le balisage normalisé des textes» et qui constitue «une norme de balisage, de notation et d’échange de corpus des documents électroniques»2. La TEI, par la publication régulière de recommandations3, propose un large panel de balises XML qu’il s’agit d’adapter aux besoins de son propre objet d’étude. Loin d’être contraignante, cette nécessité de prendre comme assise les jeux de balises et autres composants XML définis dans la TEI permet de rendre intelligibles les corpus entre chercheurs. De surcroît, si le but premier de la TEI est de permettre et de faciliter l’échange des corpus eux-mêmes, un autre atout de suivre la TEI est que cela offre une possibilité d’analyser et de confronter les approches en elles-mêmes. S’inscrire dans la logique TEI permet ainsi de s’appuyer sur l’expérience acquise dans l’élaboration des corpus de textes électroniques. Pour notre propos, notre regard doit également se porter sur les groupes de travail qui ont plus particulièrement pour objet la publication en ligne des textes épigraphiques ou en langues anciennes. Au premier rang de ces groupes de travail, il faut citer EpiDoc, ou Epigraphic Documents in TEI XML (http://epidoc.sourceforge.net/), et le Perseus Project (http://www.perseus.tufts.edu/), deux émanations de la TEI particulièrement intéressantes pour notre réflexion. L’objectif du présent article étant moins de définir la DTD propre aux textes égyptiens que de soumettre les quelques conclusions qu’il m’a paru possible de tirer de mes interrogations sur le corpus des textes des processions des temples tardifs, il s’agit désormais de se pencher plus avant sur ces textes qui s’organisent selon une structure qui sied particulièrement à un traitement XML.
1
2 3
L. Romary, H. Hudrisier, «TEI – Text Text Encoding Initiative», article en ligne sur le site Internet du RIFAL, Réseau international francophone d’aménagement linguistique, http://www.culture.gouv.fr/culture/dglf/rifal/tei.htm (dernière consultation: octobre 2008). Ibid. Nous en sommes actuellement à la cinquième édition: http://www.teic.org/Guidelines/P5/ (dernière consultation: octobre 2008).
96
INFORMATION TECHNOLOGY AND EGYPTOLOGY
2008
LES PROCESSIONS DE SOUBASSEMENT DES TEMPLES TARDIFS: HISTOIRE ET TYPOLOGIE
Élément important de la grammaire du temple, les processions de figures ornant les soubassements des temples d’époques ptolémaïque et romaine constituent un corpus dont la richesse s’avère particulièrement intéressante en ce qui concerne non seulement les études toponymiques mais également l’histoire des religions ou celle, plus générale, des mentalités. Les processions dans l’histoire Dès le début de la IVe dynastie et le règne de Snéfrou4, des personnages incarnant les provinces (spA.t) d’Égypte apparaissent dans l’ornementation des parois des temples funéraires royaux aux côtés des figurations des domaines du défunt venus lui apporter les richesses de ses terres. Les processions de personnages androgynes ou féminins seront dès lors le support privilégié pour représenter le territoire et ses richesses. Glissant du monde des défunts à celui des dieux, les défilés viendront enrichir progressivement l’iconographie des temples. Cette évolution est plus particulièrement intéressante à suivre en ce qui concerne les processions d’ordre géographique. Ce n’est cependant pas sous l’aspect d’une procession qu’apparaît la première énumération géographique connue produite dans le domaine religieux non-funéraire. C’est en effet sous la forme d’une liste disposée en tableau qu’apparaît le dénombrement de sépats que le roi Sésostris Ier a fait figurer sur le soubassement de sa Chapelle blanche à Karnak5. Plus tard, sur la Chapelle rouge érigée à Karnak par Hatchepsout – monument dont la fonction est identique à celle de la Chapelle blanche –, le tableau des sépats ornant le soubassement est remplacé par une procession de figures androgynes et féminines6. Si l’information fournie par ce collège de personnages est bien moins importante puisqu’elle ne se résume guère qu’au seul nom de la province, la charge symbolique peut être considérée comme bien plus importante puisque ce retour à la procession proprement figurée renoue
4 5 6
A. FAKHRY, The Monuments of Snefru at Dashur. II. The Valley Temple. Part I. The Temple Reliefs, Le Caire, 1961, p. 17–58. P. LACAU, H. CHEVRIER, Une chapelle blanche de Sésostris Ier, Le Caire, 1956, p. 207–251. Id., Une chapelle d’Hatchepsout à Karnak, [Le Caire], 1977, p. 69–92.
CORPUS ÉLECTRONIQUES DE L’ANCIEN ÉGYPTIEN
97
avec une forme plastique finalement beaucoup plus expressive que le tableau. Comme souvent dans l’iconographie égyptienne, le texte vient soutenir l’image, mais quoique connaissant huit variantes7, les formules accompagnant les porteurs d’offrandes de la Chapelle rouge d’Hatchepsout restent très stéréotypées et ne présentent que de manière très générique les biens présentés: «Je t’apporte toutes les choses bonnes et douces issues de la Terre du Nord», «toutes les offrandes fraîches et pures», «toutes les provisions», etc. Ce n’est qu’à partir de l’époque ramesside que la nature de ces formules s’étoffe et que celles-ci abandonnent leur aspect stéréotypé, allant parfois même jusqu’à évoquer des spécificités régionales propres8. À cette même époque, les processions présentant des figures incarnant les produits et les richesses du pays connaissent également un regain d’intérêt9. Connues par des modèles remontant à l’époque Sheshonquide10, les processions de «Nils» gravées sur les soubassements des temples d’époque ptolémaïque puis romaine vont acquérir une réelle épaisseur quant au contenu de leurs textes. Les exemples les mieux conservés sont assurément ceux des temples d’Horus à Edfou et d’Hathor à Dendara, mais rares sont les temples qui ne comportent pas de ces défilés dans leur programme décoratif (Figure 1). Formes et types des processions en œuvre sur les parois des temples gréco-romains Dans cette époque mémorielle où les hiérogrammates puisent dans leurs archives pour compiler le compendium sacré que l’on peut lire aujourd’hui, entre autres, sur les parois des derniers temples, les motivations profondes dans le choix des toponymes et des entités représentées sont complexes11. Le sujet de ces processions est tant à la fois géographique – dans le cadre d’une une géographie sacrée –, économique ou mythologique, ces trois domaines se mêlant de manière étroite. 7 8 9
10
P. LACAU, H. CHEVRIER, Une chapelle d’Hatchepsout à Karnak, [Le Caire], 1977, p. 71. J. YOYOTTE, AEPHE Ve section 98, 1989–1990, p. 179. Voir par exemple la procession du temple de Séthy Ier à Gurna (H. BRUGSCH, Die Geographie des alten Ägyptens, Geographische Inschriften altägyptischer Denkmäler 1, Leipzig, 1857, pl. XII). Id., «Note sur le bloc de Sheshonq I découvert par la mission archéologique à Saqqara de l’Université de Pisa», Egitto e Vicino Oriente 12, 1989, p. 33–35.
98
INFORMATION TECHNOLOGY AND EGYPTOLOGY
2008
En ce qui concerne les processions ornant les soubassements des temples gréco-romains, quatre grandes catégories ont été mises en évidence12: 1) les processions géographiques, lesquelles se décomposent en: a) Les défilés des seules quarante-deux figures représentant chacune une sépat d’Égypte; b) Les processions plus développées dites quadripartites, défilés ainsi dénommés parce que chaque province est représentée par quatre personnages, incarnations des quatre composantes d’une province: la spA.t elle-même, le canal-mr et les territoires w et pHw qui lui sont propres. 2) les processions hydrologiques; 3) une variante de ces dernières où sont mis en scène des couples inondation/campagne; 4) les processions économiques. Structure générale des textes des processions de soubassement Quoique des variations en nuancent l’apparente unicité, ainsi que le montre la typologie rappelée ci-dessus, les textes des processions forment un ensemble cohérent tant du point de vue de la forme que du contenu, ce qui rend particulièrement intéressante leur mise en corpus. Ainsi, en prenant en considération les seules processions géographiques par exemple, un rapide examen permet de voir que les textes de chaque province s’articulent autour de deux syntagmes13: un premier présentant l’entité de la procession et les produits qu’elle apporte avec elle; un second définissant un certain nombre d’assimilations divines destinées à caractériser la divinité récipiendaire en fonction de l’apport (la tradition égyptologique désigne cette seconde partie sous le nom de «glose d’assimilation»). À l’intérieur de ces deux syntagmes, plusieurs composantes grammaticales et sémantiques donnent corps à ce système. Si l’on prend par exemple le texte de la procession concernant la XIXe sépat de Basse-Égypte, il est possible de mettre en exergue différents éléments tel que cela est illustré dans la Figure 2. 11
12 13
Voir J. YOYOTTE, Orientalia 35, 1966, p. 46; id., AnnEPHE Ve section 75, 1967– 1968, p. 106; 91, 1982–1983, p. 217–221; Chr. ZIVIE-COCHE, Tanis. Statues et autobiographies de dignitaires. Tanis à l’époque ptolémaïque, TTR 3, Paris, 2004, p. 294; V. RAZANAJAO, « Le Delta à Basse époque: géographies d’un territoire», Égypte. Afrique et Orient 42, juin 2006, p. 3–9. J. YOYOTTE, Annuaire du Collège de France 94, 1993–1994, p. 685–686. Dans le sens large donné à ce mot par F. de Saussure, Cours de linguistique générale, Lausanne, Paris, 1916, p. 172.
CORPUS ÉLECTRONIQUES DE L’ANCIEN ÉGYPTIEN
99
Syntagme d’assimilation
Syntagme de présentation
Figure 1: Edfou, soubassement du mur ouest de la cour du temple d’Horus. Début de la procession quadripartite des sépats de Basse-Égypte © Photo de l’auteur. Verbe de l’apport suivi du pronom désignant le roi
Prép. du datif + pronom désignant la divinité récip.
Entité apportée
Préposition introduisant l’offrande
Produits apportés par le personnage
Toponyme
Toponyme
in≠f n≠k Jm.t-pH Xr mXr.w≠s m jrp n(y) TA-nTr Il t’apporte Imet-peh chargée de ses produits qui sont le vin de Ta-netjer Ntk jm nTrj xnty wnm.t≠f sD.tj wr xnty jAb.t≠f Car tu es l’enfant-imti divin qui préside à son œil-droit, l’enfant-sdjti vénérable qui préside à son œil gauche
Pronom désignant la divinité récipiendaire
Assimilations dont la divinité récipiendaire est l’objet
Figure 2: Mise en exergue des éléments composant les textes accompagnant une figure de procession, en l’occurrence la sépat de la XIXe province de Basse-Égypte (Edfou IV, 37, sqq.)
100
INFORMATION TECHNOLOGY AND EGYPTOLOGY
2008
Un traitement selon les normes XML du texte de la Figure 2 consisterait à indiquer en premier lieu qu’il est question de la procession ornant le soubassement du pronaos du temple d’Edfou (= Edfou IV, 21–39 pour la Basse-Égypte; IV, 172–193 pour la Haute-Égypte). À un niveau inférieur, il faudrait renseigner le fait qu’il s’agit du texte de la XIXe sépat (Edfou IV, 37, 3–38, 1) puis que de «in=f» à «TA-nTr», nous avons affaire au syntagme de présentation, et de «ntk» à «jAb.t=f», au syntagme d’assimilation. Par la suite, il est possible de marquer les différents éléments constitutifs du texte, qu’il s’agisse d’objets grammaticaux (préposition introduisant l’apport, pronom introduisant la divinité récipiendaire) ou sémantique (assimilations divines dont la divinité récipiendaire est l’objet). Un balisage XML permettrait également de renseigner le type de personnage qui incarne l’entité apportée, celui de sa coiffe, sa position dans la procession. D’un caractère plus «flottant», un autre balisage aura son importance: celui des toponymes. Pouvant apparaître dans les deux syntagmes, les noms de lieux pourront ainsi être marqués, quelle que soit leur position dans le texte. La Figure 3 donne l’ébauche de la structure générique des processions qu’il paraît possible de mettre en avant. Il reste à s’interroger sur la manière que l’on pourrait user pour traduire en XML ce schéma générique, et quels éléments et balises de la TEI il s’agit de prendre en considération. POUR UNE DFINITION D’UNE DTD-TEI PROPRE AUX TEXTES ÉGYPTIENS
Un premier jet de DTD liée aux textes hiéroglyphiques a vu le jour en 200014. Comme son auteur S. Rosmorduc l’explique en introduction, cette DTD n’avait pas pour vocation de baliser les textes hiéroglyphiques d’un point de vue sémantique mais bien du seul point de vue formel. La question était centrée sur les problèmes que pose la saisie même des textes hiéroglyphiques et la manière de les encoder, indépendamment presque du XML, suivant les préceptes du Manuel de codage. Mais, et particulièrement dans le cadre qui nous occupe ici – celui des textes des processions de soubassement –, il s’avère que l’intérêt du XML et de l’application des 14
Consultable sur les pages Internet de S. Rosmorduc à l’adresse http://webperso.iut.univ-paris8.fr/~rosmord/HieroEncoding/DTD/ (dernière consultation: octobre 2008).
CORPUS ÉLECTRONIQUES DE L’ANCIEN ÉGYPTIEN
101
Procession
Sépat
Sépat
Figure Sexe Numéro d’ordre Pavois
Sépat
Sépat
Sépat
Texte Syntagme de présentation Verbe introductif
Syntagme d’assimilation Vocabulaire
Prép. datif + pron. div.
Toponyme
Entité apportée
Toponyme
Produits apportés
Toponyme
Pronom introductif Assimilations Assimilation Assimilation
Figure 3: Arborescence de la structure générique applicable aux processions de soubassement
recommandations TEI réside dans les réponses que ceux-ci apportent à la façon dont on peut appréhender un corpus textuel sur le plan sémantique. Malgré les apparentes dissemblances qui séparent les textes égyptiens des manuscrits médiévaux ou modernes – principal sujet d’études des corpus TEI –, il semble possible d’affirmer que l’ensemble des entités et balises contenues dans la TEI répondent aux besoins, à l’instar de ce qui a été fait dans le cadre du projet EPIDOC. Aucune création de balise n’est nécessaire car l’adaptation de celles existantes par le biais des définitions d’attributs est très souple et extensible. Dès l’abord, il apparaît que les entités définies par la TEI afin de renseigner les informations générales sur un manuscrit et destinées à figurer dans l’en-tête fonctionnent parfaitement. Le «header» du fichier XML sur les processions pourra donc reprendre les éléments de la TEI suivant:
( file description) contient la description du fichier électronique d’un point de vue bibliographique.
102
INFORMATION TECHNOLOGY AND EGYPTOLOGY
2008
(text-profile description) contient la description du texte contenu dans le fichier électronique du point de vue de son élaboration, de son contexte de production, de sa langue (balises , , etc.). (classification declarations) répertorie les références qui seront utilisées pour des renvois externes au document. Par exemple, il est possible de définir les abréviations bibliographiques qui seront utilisées tout au long du corpus (Dendara, Edfou, Wb, PM, etc.). Afin de baliser la structure générale de chaque texte relatif à une figure, des avec trois «types» différents pourraient être utilisées: , division concernant les informations sur la figure elle-même (genre, numéro d’ordre, pavois); , division concernant les textes du syntagme de présentation; , division concernant les textes du syntagme d’assimilation; À l’intérieur de chacune de ces divisions, il s’agit de pouvoir saisir des hiéroglyphes, de la translittération et de la traduction. Les suivantes pourraient être utilisées: pour les hiéroglyphes; pour la translittération («al» = «alphabetic»15); pour la traduction. Ces dernières étant d’un niveau inférieur à celui des , etc., l’utilisation de «div» hiérarchisées sera un avantage. Ainsi, pour les grandes parties de texte, pour les séquences hiéroglyphiques, de translittération et de traduction. L’adaptation de l’existant permet également de ne pas avoir à créer de balises particulières pour l’encodage des signes hiéroglyphiques. Il nous paraît ainsi envisageable d’utiliser les balises suivantes:
(character or glyph) pour le balisage d’un signe hiéroglyphique unique.
15
Les suffixes « hi », « al » et « tr » sont repris de M. J. NEDERHOF, « Alignment of Resources of Ancient Egyptian Texts Based on XML », dans Proceedings of the 14th Table Ronde Informatique et égyptologie, Pise, 2002, ressource en ligne: http://www.cs.st-andrews.ac.uk/~mjn/egyptian/align/index.html, p. 2 (dernière consultation: octobre 2008).
CORPUS ÉLECTRONIQUES DE L’ANCIEN ÉGYPTIEN
103
(segment) permettrait de procéder aux regroupements de signes. Une définition adéquate de l’attribut type permettrait d’indiquer le type de regroupement: cadrat ou, à l’intérieur d’un cadrat, groupe horizontal, vertical, ou encore ligature. e.g.: W25A9 nk
A18pH O49
L’indication de hachure pourrait être portée sur un texte par l’utilisation de la balise dont le type serait défini comme «hatch». Les balises TEI , , , , le couple / (ou la balise ?) et permettront de porter toutes les indications utiles à l’édition du texte. En ce qui concerne la translittération et la traduction, diverses balises permettront d’affiner l’encodage sémantique des mots et séquences qui composent le texte:
(word ) ainsi que la définit la TEI, «represents a grammatical (not necessarily orthographic) word». Cette balise permettra par exemple de marquer un toponyme. L’attribut «lemma» sera utile pour préciser la forme exacte du mot. e.g.: le balisage de la translittération de pourrait être rendu par: Jm.t-pH
(referencing string),
permettrait de marquer, par exemple, que la partie du texte ainsi balisée concerne les produits apportés. e.g.: cf. texte du syntagme d’assimilation de la Figure 2. ntk jm nTrj xnty wnm.t≠f sD.tj wr xnty jAb.t≠f Comme on peut le voir, les préceptes établis par la TEI répondent aux besoins de l’égyptologue en matière de traitement d’un corpus textuel. Je ne saurais prétendre, par cet article, établir une DTD-TEI propre aux textes égyptiens, ceci nécessitant un évident travail en commun. Aussi, je n’approfondirai pas plus la description des adaptations qu’il m’a paru possible de mener sur les balises XML-TEI. J’ai toutefois annexé en fin d’article une ébauche de fichier XML qui permet, je l’espère, d’illustrer la manière dont il paraît possible d’aborder les textes égyptiens.
104
INFORMATION TECHNOLOGY AND EGYPTOLOGY
2008
Les réflexions dont je vous ai fait part ont été induites par la nature même du corpus que je souhaitais traiter, celui des textes des processions de soubassement des temples tardifs. Cet article se veut avant tout l’annonce de ce projet et une pierre dans l’établissement d’une DTD pour les textes en ancien égyptien. Il apparaît nécessaire, pour l’avenir, d’établir un groupe de travail à l’image d’EPIDOC, groupe de travail égyptologique qui aurait sa place toute trouvée dans le cadre des tables rondes Égyptologie & informatique. ANNEXE. ÉBAUCHE DE FICHIER XML-TEI POUR L’ENCODAGE DES TEXTES DES PROCESSIONS DE SOUBASSEMENT DES TEMPLES TARDIFS En-tête de fichier avec le titre du corpus dans son entier; Filedesc: description Corpus des textes des processions ornant les soubassements des générale du fichier XML, temples ptolémaïques et romains
avec indication de l’éditeur
(respStmt);
Corpus of procession-texts inscribed on the wall-basement of Egyptian Temples of Ptolemaic and Roman Periods
Edited by Édité par Vincent Razanajao
Édition des textes des processions ornant les soubassements des temples égyptiens d’époques grecque et romaine
Publication of the texts of processions inscribed on lower parts of walls in Egyptian temples of Greek and Roman date
Textes relatifs à la première sépat de Basse-Égypte. Textes concernant les sépats 2 à 18 Début de la section concernant les textes relatifs à la XIXe sépat de BasseÉgypte. Header avec les indications bibliographiques
Edfou IV, 37, 3-38,1
PM VI, 157, (290)-(294)