The process of reading a fragmentary ancient document in any medium has several facets. Although some ancient texts can be read easily, many are impenetrable to the naked eye. This is frequently due to damage, fading of characters, multiple handwriting, incomplete or confusing strokes or the palimpsest nature of the document (two or more texts written one on top of another). The script itself must therefore be reconstructed, individual letters, words and groups of words ascertained, and meanings and translations attached to those words. The project Image, Text, Interpretation: e-Science, Technology and Documents, also known as eSAD: e-Science and Ancient Documents, is based at the Centre for the Study of Ancient Documents at the University of Oxford, UK. The project collaborates with Computer Scientists, seeking to provide e-science tools and methods which may facilitate the decipherment and interpretation of damaged documents, such as the writing ink and stylus tablets from the Roman fort of Vindolanda, few of which have survived intact. The project is developing a number of digital tools to be made available to epigraphers, papyrologists and historians, offering means of tracking the interpretative process and the development of hypotheses together with image processing functionality.
This case study describes work in progress and will be updated as the project evolves.
Fig 1. Encoding a Vindolanda ink tablet. © eSAD Project. Source: Building a Decision Support System, Presentation by Henriette Roued Olsen, p. 18, available at http://esad.classics.ox.ac.uk/index.php?option=com_docman&Itemid=96
Description of aims
Description of methods
Description of outcomes and issues
Details of any training provided
Technologies, tools, software, services and resources used
Standards employed; data formats and protocols
Significance for Research
Future Work Enabled
Evidence of Dissemination Activities
Conclusions about e-Science
Further information, bibliography and links
Email: through the 'Contact us' page at http://esad.classics.ox.ac.uk/
The eSAD project aims to develop and deliver digital tools and methodology which may support reading ancient documents on papyri, wood and other supports, particularly those which are damaged and open to ambiguous interpretation of the text. It seeks to provide experts with an interpretation support tool employing tailored image processing technology, which does not require a high level of technical skills and that may be used routinely and in accordance with the established epigraphical and linguistic conventions. The project therefore explores the use of a Decision Support System (DSS) in the day-to-day reading of ancient documents for keeping a record of how the documents are interpreted and read, thus making this complex process transparent. This approach requires a good understanding of how epigraphers work with text and how this can be translated into a machine environment; how they deal with lack of evidence and how the reconstructing of the original meaning may be supported through computational means. A combination of image processing tools and an ontology-based support system are being developed to facilitate experts by tracking their developing hypotheses. The system will offer suggestions of possible alternative readings based on the analysis of patterns in linguistic and palaeographic data. The project also aims to investigate how the resulting images, image tools, and data sets can be shared between scholars within a virtual research environment.
The project is developing new methods and unique tools, as well as building upon earlier research, drawing on both traditional and digital scholarship, by using and enhancing existing resources, and through collaboration with other projects and experts, details of which are given below.
Inscribed (stylus) tablets
Fig. 2 Left. A Latin writing tablet from Tolsum in Frisia. © eSAD Project. Source: http://esad.classics.ox.ac.uk/images/stories/m2c_frisia-005061.gif
Fig. 2 Right. The same tablet after a woodgrain removal algorithm was applied to it. © eSAD Project. Source: http://esad.classics.ox.ac.uk/images/stories/roi_frisia-005061.gif
The ancient Latin documents serving as main test set for this project include stylus tablets which were found, from 1973 onwards, on the site of the Roman fort of Vindolanda. This important military post was located on the northern frontier of Britain, and pre-dates the building of Hadrian's wall. The wooden tablets from Vindolanda are now in the British Museum. They are wafer thin and fragile, and are inscribed in Roman cursive script either in carbon ink directly on the wood or incised with a stylus in wax-coated tablets, of which the wax has almost always disappeared leaving only the incisions below it on the wood. Approximate dimensions for one complete leaf are 12cm x 7cm. The tablets date from around AD 100 and are the oldest surviving handwritten documents in Britain. They offer invaluable insights into the military life at Vindolanda, and about pleasures and ceremonies, clothing and food, administrative, financial and other matters. The tablets were found in deposits of rubbish and mud which had a deleterious effect on their condition. Many have survived only in fragments and the reconstruction of the abraded text is often problematic. The tablets have been published (transcribed and translated from Latin into English) in an electronic resource, Vindolanda Tablets Online which is illustrated and offers flexible ways of querying the contextual information in the documents. The content and functionality of this resource has been enhanced by the doctoral student affiliated to the eSAD Project, Henriette Roued-Cunliffe. The Vindolanda Tablets Online now supersedes earlier editions of 1983, 1994 and 2003 by Alan Bowman and David Thomas published in book format.
A number of imaging techniques have been applied to capture inscriptions in much detail and in particular to provide a variety of views of the problematic areas, to enable complex examination. A technique similar to Polynomial Texture Mapping (PTM) which has also been used on paintings and other artefacts, was used by the eSAD project in the summer of 2008 to capture images of the Vindolanda tablets in the British Museum, in collaboration with Dr Graeme Earl of the University of Southampton. Dr Charles Crowther also carried out considerable image capture at the Centre for the Study of Ancient Documents, Oxford, for lead tablets and stylus tablets. This technique uses the shadow stereo principle and involves photographing the artefact by capturing a series of images of the object's surface; the angle of projected light is changed with each image, while the vantage point is kept fixed. The resulting composite image captures the three-dimensional nature of the object surface and textures.
The published editions of the Vindolanda ink tablets have been encoded using XML mark-up thus allowing for a detailed and automated indexing of these. The encoding has covered contextual items (e.g. people or military terms), which now supplies a knowledge base of words used in the Vindolanda documents, through a web service. This knowledge base can be used as a stand-alone resource but will be integrated into the Decision Support System described below.
Image processing techniques are being used to isolate the text from the background for incised tablets. A set of physical context correction algorithms was developed to tackle woodgrain removal (using Lambertian reflectance compensation, see Fig. 2) and for illumination correction (through homeomorphic filtering). A stroke detection algorithm is currently (June 2009) under development. It involves Fourier analysis, computation of the monogenic signal associated to the image to determine phase congruency. A new approach to measuring the actual phase congruency as a method to detect the strokes is also under development.
The incisions detected by the image processing algorithms then constitute candidates for strokelets that are part of character strokes. Based on the samples of character shapes already known, each character is described as a set of strokelets with particular orientations and positions with respect to one another, allowing a simple geometrical description of the palaeographical data. These descriptions allow for identification of possible letters as well as integration of newly found occurrences of letters. Variations in grouping of strokelets yield different possibilities for letter identification (Fig 3).
Fig 3. Example of an evidence based DSS. © eSAD Project. Source: Building a Decision Support System, Presentation by Henriette Roued Olsen, p. 11, available at http://esad.classics.ox.ac.uk/index.php?option=com_docman&Itemid=96
The foundations of a Decision Support System (DSS) have been set. This aspect of research requires good understanding of experts' behaviour when reading damaged ancient texts in order to model this process computationally. Building on Melissa Terras’s earlier work about the process of reading ancient documents, the DSS will support the reader through this difficult process on the reader’s own premises. The system is based on an interpretative tree model and accommodates possible alternative readings. A percept-based approach has been adopted, where a percept is the result of an interpretation to some level. The aim of the DSS is to build a network of substantiated percepts exposing the cognitive processes that are involved in the interpretation task and that may yield one or more (possibly conflicting) interpretations.
The DSS builds on an idea of evidence based interpretations. Each percept or minor interpretation will be either supported or opposed by pieces of evidence. The knowledge base gives a list of words, which can be used to support or oppose a percept of a word. Evidence of type 'judgements' (e.g. 'because I say so!' in Fig. 3) will also be allowed, thus leaving full control of the tool to the user. An elementary percept is defined as a tile of an image containing what is perceived to be a character; thus linking up to the image processing aspect of the project through the geometrical description of characters and grouping of strokes with the DSS. The project team is working on identifying higher level percepts (which progressively add meaning to the observed signal, turning the signals into symbols); and on building an ontology to support reasoning, and further, reasoning under uncertainty.
Fig 4. Example of two groupings of strokelets yielding two different interpretations. © eSAD Project. Source: Image Capture and Analysis, Presentation by Ségolène Tarte, p. 23, available at http://esad.classics.ox.ac.uk/index.php?option=com_docman&Itemid=96
Although some ancient texts can be read easily, many are impenetrable to the naked eye. This is frequently due to damage, fading of characters, multiple handwriting, incomplete or confusing strokes or the palimpsest nature of the document (two or more texts written one on top of another). Digital imaging and other computer techniques provide tools for decipherment of text through enhancement of features and removal of noise.
In the first year the project has explored image processing techniques; mark-up and web-service technology for building and accessing knowledge bases; use of e-infrastructure; and methodologies to identify mechanisms in deploying implicit knowledge and build an ontology; all of these converging towards the project's final aim, the construction of a web-based interpretation support system for papyrologists and epigraphers.
Digital tools developed by eSAD have been integrated into a prototype Virtual Research Environment (BVREH). This integration work was carried out by Pin Hu, Kang Tang, Ségolène Tarte and David Wallom of the Oxford eResearch Centre, University of Oxford, as part of the ENGAGE initiative and supported by the Open Middleware Infrastructure Institute (OMII-UK), University of Southampton. The project Building a Virtual Research Environment for the Humanities (BVREH), funded by the UK Joint Information Systems Committee (JISC) was developed by Pin Hu, Ruth Kirkham, John Pybus (OeRC) , Dr Charles Crowther and Prof. Alan Bowman.
Between January-October 2008 Henriette Roued Olsen was in training at the Centre for the Study of Ancient Documents of the University of Oxford working on advanced XML mark-up and generation of a knowledge base of the Vindolanda ink tablets.
XML,XSLT, RESTful web services, digital imaging, image processing with Matlab Polynomial Texture Mapping (PTM), web services.
eSAD operates under Open Source principles. EpiDoC standard XML and RESTful web services protocol are used for the building of the knowledge base. JSR-168 compliant portlets, GridSAM, NGS resources for the integration with the VRE.
The project, although still in its early stage, has already made a significant contribution to scholarship in the area of epigraphy by correcting some of the earlier readings of the Vindolanda tablets. An integrated approach to digital tools and methodology (in progress) adopted for eSAD and the use of existing e-science infrastructure are enhancing the day-to-day practice of researchers and continue pushing interdisciplinary barriers. Of particular significance is the contribution made by the experts in radiology and breast cancer research. The processes of decision making in medical diagnosis is informing eSAD. Full significance of the project will be evaluated in due course.
This will be reported upon towards the end of the project.
8-11 September 2008, The UK e-Science All Hands Meeting: Crossing Boundaries: Computational Science, E-Science and Global E-Infrastructures, University of Edinburgh, Scotland.
13 May 2009, Image, Text, Interpretation: e-Science, Technology and Documents, A workshop presenting the results of the work completed in the first year of the project. Oxford e-Research Centre, University of Oxford, UK. Selected presentations are available from the event page at http://esad.classics.ox.ac.uk/index.php?option=com_docman&Itemid=96.
These will be provided towards the end of the project.
Project website: http://esad.classics.ox.ac.uk/
Vindolanda Tablets Online http://vindolanda.csad.ox.ac.uk/exhibition/
Alan Bowman and David Thomas, Vindolanda: the Latin writing tablets, vol. i, Society for the Promotion of Roman Studies, 1983; The Vindolanda Writing Tablets (Tabulae Vindolandenses), vol. ii, London: British Museum Press, 1994; Addenda and Corrigenda, vol. iii, 2003.
Building a Virtual Research Environment for the Humanities Project (BVREH), June 2005-September 2006, http://bvreh.humanities.ox.ac.uk/
The Virtual Environment for Research in Archaeology Project (VERA), http://vera.rdg.ac.uk
Alan K. Bowman, Charles V. Crowther, Ruth Kirkham and John Pybus, A Virtual Research Environment for the Study of Documents and Manuscripts, Paper presented at the conference Oxford eResearch 2008, 11-13 September 2008, University of Oxford, available at http://ora.ouls.ox.ac.uk/objects/uuid%3Ad7f250e0-9a95-4193-b476-8666ce5c3347.
ENGAGE Project, http://www.engage.ac.uk
Information on the eSAD Project on the website of the Open Middleware Infrastructure Institute (OMII-UK), University of Southampton:
A view on ancient documents', OMII-UK News, March 2009, http://www.omii.ac.uk/wiki/Nwsltr0309eSAD
Ségolène Tarte, 'Image Acquisition & Analysis to Enhance the Legibility of Ancient Texts', Presentation [no. 1030] from the UK e-Science All Hands Meeting, Crossing Boundaries: Computational Science, E-Science and Global E-Infrastructures held at the University of Edinburgh between 8 and 11 September 2008 is available at http://www.allhands.org.uk/archive/2008/programme/index.html; an extended abstract is available on the eSAD project website, http://esad.classics.ox.ac.uk/
This case study is based on the project interim report of December 2008, the project website and material from related events. It was compiled for the Arts and Humanities e-Science Support Centre (AHeSSC), King’s College London, by Anna Bentkowska-Kafel, 18 May 2009. Aditional information kindly provided by Henriette Roued Olsen and Ségolène Tarte.
© e-SAD Project, June 2009.Back to e-Science Initiative projects page