On the basis of a comparison of several encoding options provided by the TEI-P5, this paper presents a TEI-conformant basic encoding scheme for the representation of selected genres of computer-mediated communication (CMC).
The authors of this paper are, on the one hand, corpus providers (cf. paper 2 in this panel) and, on the other hand, linguists who pursue corpus-based research on digital genres and on language use on the internet (cf. paper 1 in this panel). The encoding scheme should therefore not only meet the requirements of corpus building and integration but should also enable linguists to annotate and analyze the particular linguistic and structural properties of CMC discourse.
Thus, a TEI schema for CMC discourse that serves both purposes equally well would have to fulfill the following requirements that are adressed with a little more detail in the complete version of this abstract which can be found in the attached PDF file:
In order to find out which encoding best meets the requirements given above, we are experimenting with several encoding options which are based on modules that are provided by the current version of the TEI guidelines: the module 4 “default text structure” (which provides a model for a broad range of text types which have been produced and edited under monologic conditions), the module 8 “transcriptions of speech” (which provides a model for spoken dialogues) and the module 7 “performance texts” (which provides a model for written dialogues – with individual speeches assigned to several speakers – but whose concept does not comprise natural conversations).
We investigate to which extent these modules can be customized for CMC discourse. In our presentation, we will report these experiments and compare the merits and drawbacks of the different encoding options in light of the above-mentioned requirements. As a result, we will present an encoding scheme which fits well with the framework of the DWDS corpus and which makes a compromise between the requirements outlined above and the modules that are provided by the current version of the TEI.
In our examples, we will focus (1) on threads in online forums/discussion boards, on discussion pages of Wikis and on “social network” sites (asynchronous mode), and (2) on logfiles from IRC, webchat and instant messaging conversations (synchronous mode).
The encoding options as well as our suggestions for a basic TEI encoding scheme for CMC discourse will be illustrated using examples from a dataset that we retrieved from the internet in 2010 and 2011 and which includes data from a broad range of genres.
The paper will finish with an outlook on features of CMC discourse that are not yet covered by the presented format and how they could be integrated into the current annotation framework. We would like the audience to discuss the presented basic CMC encoding scheme as well as the formulated desiderata.
The panel addresses issues related to a project that aims at building a TEI-compliant reference corpus of German computer-mediated communication (CMC). This corpus (‘Deutsches Referenzkorpus zur internetbasierten Kommunikation’, DeRiK) will cover a broad range of CMC genres such as e-mail, discussion boards, chats and instant messaging conversations, weblogs, wiki discussions, microblogging on Twitter and communication in “social network” sites. It shall be integrated into the DWDS corpus collection of contemporary German (Geyken 2007) and used in the context of corpus-based lexicography. In our panel we want to present data from ongoing work in our project, compare several modeling options which apply different TEI modules and discuss a draft for a TEI-based core format for the representation of CMC genres. The overall goal of the panel is to pave the way for a TEI-based encoding scheme for CMC discourse which meets the specific requirements in the DeRiK context but which may also be used by other projects that aim at building annotated corpora of CMC.
Up to now, corpus-based CMC projects have typically developed their own, project-specific encoding schemes. This complicates if not even inhibits the sharing of the data across projects. This is all the more regrettable because many projects add value to the data through their annotation. Sharing, merging and comparing datasets, particularly in constrastive research, calls for a standard-conformant basic scheme which suits the need of various projects and which is easy to handle and extend. Since many resources within the humanities use the TEI framework for annotation purposes, such a basic scheme for CMC should be conformant with TEI.
The papers of the panel will point out crucial challenges of building CMC corpora from the perspective of linguistic CMC research, discuss several options for the representation of CMC data on the basis of customized modules of the TEI-P5 and present a draft for a TEI-based encoding schema for the representation of CMC. The panel is structured as follows:
The panel should be concluded by a discussion of the basic schema for CMC presented in paper 3.
The paper gives an outline of the essential challenges of creating a TEI encoding scheme that captures the structure and properties of computer-mediated communication (CMC). From the perspective of linguistic CMC research and with the help of examples, we will point out that the discussion about a basic format for the representation of CMC should carefully reflect the following issues (cf. Beißwenger & Storrer 2008) which are outlined in detail in the complete version of this abstract which can be found in the PDF attachment:
On the basis of outlining these issues, the paper discusses the general decisions with which one is faced when aiming at modelling CMC data using the framework provided by the TEI-P5. These general decisions are related to questions such as whether CMC discourse could be adequately represented in terms of the TEI-modules for (a) transcriptions of speech (which provides a model for spoken and not for written language), (b) text (which provides a model for a broad range of text structures which have been produced and edited under monologic conditions) or (c) performance texts (which provides a model for written dialogues – with individual speeches assigned to speakers – but whose concept does not comprise natural conversations), or whether CMC discourse, due to its crucial differences to any of the genres already recognized by the TEI-P5, should rather be treated in an own (not yet existing) module of a future version of the TEI framework.
The general issues discussed in this paper will be taken up in paper 3 and readdressed using data from several CMC genres out of the context of the DeRiK project (cf. paper 2) as well as encoding examples for these data applying several modules provided by the TEI-P5.
Les curieuses recherches du Mont Saint Michel have been written by dom Thomas Le Roy and are the object of an electronic edition (work in progress). The monk wrote, for the two years he spend in the abbey of the Mont Saint-Michel (november 1646-july1647), three texts about the history of the abbey, under three different forms : an abstract of 20 pages ; a thematic texte of 200 pages and a chronological text of 600 pages. He used then of the rich library of the abbey, and follows, at least for a part, the recommendations of his congregation, the congregation of Saint-Maur.
I will demonstrate how the Text Encoding Initiative, which provides guidelines on structuring humanities and social science texts in XML (eXtensible Markup language) has been applied to the specifities of the corpus of dom Thomas Le Roy and also the project objectives. This edition should not only allow an entire version of the corpus to be consulted (until now, dom Thomas Le Roy’s work had never been published in its entirety), but provides an opportunity to analyse the work of monks in the context of the Maurists historical reform in the XVIIe and XVIIIe centuries.
Although still a work in progress, the encoding of three manuscripts (BNF 13818 – twenty page manuscript written by dom Le Roy and sent to the abbey of Saint-Germain-des-Prés in July 1647 ; BNF 18950 – two hundred pages sent to the abbey of Saint-Germain-des-Prés in July 1648 ; fonds Mancel 195 – six hundred pages kept at the Abbey of the Mont Saint-Michel) allows us at this stage a glimpse of potential avenues of more precise research. This project allows a perspective on the interpretative potential that encoding presents for the corpus as it stands.
I will speak about my methodology, showing how I am using the TEI and what element have been chosen, in regards of my hypothesis (those of the beginning and those which have appeared since). I will explain how the TEI suits my corpus in three points : normalisation, analysis and publication. I will talk about the inventory of the element I am using. I will show what tools, the XML allows me to use for the study and the analysis of my corpus (stylesheets, software, transformation tools…).
The edition of Goethe's “Faust” (https://faustedition.uni-wuerzburg.de) will provide all relevant manuscripts of “Faust” by making the facsimiles and transcriptions available. It is the aim of the edition not only to represent these manuscripts as manuscripts, but to reconstruct and visualize their genetic relations. Therefore we seek to develop new visualization strategies for the electronic medium. The user of the edition will be able to follow the genesis of one of the most important literary works in the German language in every particular detail. The edition will provide a text of “Faust” including all its drafts and an exhaustive account of the totality of the textual variances. For the encoding of the transcripts, the markup of the TEI is used. The encoding model for Genetic Editions as developed by the TEI Special Interest Group on Manuscripts (TEI MS SIG) developed an (http://www.tei-c.org/Activities/Council/Working/tcw19.html) is of substantial importance to our work.
This contribution will present how the TEI markup is used in the edition of Goethe's “Faust”. Every manuscript is considered under two different perspectives: first as a piece of paper containing a material inscription and second as a medium of an abstract object, the text.
Under the first perspective the manuscript will be represented in a 'diplomatic' transcript of its inscription and in a visualization of the material structure of the bundle of papers to which a particular manuscript page either still belongs or did once belong.
Under the second perspective the representation is not necessarily limited to single manuscripts, even if the first step will always be to give a textual transcript of the inscription of single manuscripts. Furthermore, the genesis of the drama took over 60 years, and this process did not only affect the verbal shape of the work, but also its core, that is to say the conception of the work.
At the beginning, these two perspectives were intended to be combined in one complex encoding procedure within a single data file. But very soon, the fact had to be acknowledged that a combination of two widely differing perspectives leads to serious problems, especially the problem of overlapping hierarchies and/or the problem of being continuously obliged to use elements that are mutually exclusive. A special case in point is the rearrangement of segments by changing their respective positions on the manuscript.
Only the most radical solution proved to be manageable, the separation of the two levels (inscription and text) by using two separate data files. Now, both markups – the one for rendering the inscriptional record and the other one for encoding the text – can be applied without any conflict. However, the separation of the two levels on the other side does not imply that both are meant to be completely independent from each other. On the contrary, the interrelationship between both levels is of great importance for the genetic analysis.
This way of transcribing is exactly the TEI conformant model of multiple encodings of the same information described in the TEI guidelines (see chapter 20.1: www.tei-c.org/release/doc/tei-p5-doc/en/html/NH.html. As every method it has its advantages and disadvantages. The advantages, especially for dealing with Faust manuscripts, will be explained in our presentation. The splitting of the encoding rules in two different bodies of rules and concomitantly the division of most of the markup into two different types of markup shall be summarized. It will be illustrated with examples that will give some more insight into the asserted necessity of distinguishing the two perspectives.
And finally there are many questions of how to deal with the obvious disadvantages that come along with the division of transcripts, that is to say the “the maintenance of multiple copies of identical textual content” as well as the the missing “explicit indication that the various views, which might be in separate files, are related to each other” (http://www.tei-c.org/release/doc/tei-p5-doc/en/html/NH.html#NHME). What are the practical consequences of separating the diplomatic from the textual transcript for further steps in generating the edition and how will it be possible to avoid inconsistency? How are we going to evaluate and how to relate all information distributed on both levels? What do we have to keep in mind for the implementation of the genetic reconstruction?
The TEI plays an outstanding role as a first approach towards a standardized representation of annotations. While the standardization of the representation (encoding) of annotations has steadily evolved over the last years, the presentation or visualization of markup and its implications for the usability of markup tools has been treated with significantly less attention. In many cases the representation of markup and its actual presentation to the user are but the same thing: plain text (original text) and markup tags (annotation) at the code level. Peter Flynn has observed that “markup experts” have a different idea about the structure of a text than “conventional writers”, the one group seeing a document as a hierarchical tree with different kinds of nodes, and the other group seeing a text as a “continuous linear narrative, broken into successive divisions” (Flynn 2009). This divide between markup experts and plain annotators is of particular importance for the case of the TEI, which was designed as a representation standard for the humanities, social sciences and linguistics. Due to the research tradition and prevailing methods in this field, humanists often lack deep technical skills, i.e. many of them aren’t aware of basic markup concepts such as document types or document trees. At the same time, it has become clear that tool and ICT usage is as indispensable for the humanities as for any other field of research (Toms & O’Brien 2008). Santos & Frankenberg-Garcia claim that “most existing corpora today are only available to and understood by a small, restricted community of users” (Santos & Frankenberg-Garcia 2007). This makes usability and user experience on the presentation side a vital component for markup- and analysis-tools. Consequently, several tools try to hide the actual representation of markup from the user, by providing different interface designs and visualizations of the data. Unfortunately, current approaches to these issues are often not in accordance with existing usability standards like ISO 9241-110:2006 for dialog principles or ISO 9241-151:2008 for the usability of web interfaces (Dipper et al. 2004, Burghardt & Wolff 2009).
We identify two major challenges for the presentation layer of markup software which should be considered by tool designers in order to enhance the acceptance and the actual usage of standardized markup like the TEI guidelines, and to prevent corpora from becoming expensive data graveyards (Soehn et al. 2008), as corpus creation and especially intellectual annotation are extremely cost- and labor-intensive tasks. These challenges are at the same time general and domain-independent requirements for tools which strive for a high level of usability and user experience. The first requirement is an adequate visualization of data and annotation, the second requirement calls for appropriate interface and interaction design for markup- and analysis-tools. These requirements affect different stages in the typical workflow for the creation and use of corpora, which we call corpus pipeline. The corpus pipeline describes all steps necessary to fulfill an information need by querying an annotated corpus, starting from the creation and annotation of the actual corpus and ending with the query building and visualization of results. As presentation and representation often can’t be separated precisely during the first two stages (“digitization” and “normalization”) the presentation of markup is mainly an issue in the succeeding stages: “annotation”, “query building” and “visualization of results”. In the paper, we will derive and explain specific requirements for the presentation and interaction layer of each of these three stages (e.g. visualization of original text and multiple layers of annotation as well as the underlying annotation scheme during the annotation stage) by looking at existing tool solutions and by comparing different user interface design models with each other (recent examples of visually enriched tools including e.g. WordTree (Wattenberg & Viegas 2008) and DoubleTree (Culy & Lyding 2010) in the stage of “visualization of concordances”).
We argue for a user-centered presentation of markup, starting from the annotation of text, and ending with the querying of a corpus of documents and the presentation of query results. Future work will include a detailed user study and evaluation of different presentation aspects, such as the best presentation of multilayer annotation or complex queries.
This paper takes as its starting point a music theory text known as the *Ars nova*. This text has been considered foundational to our understanding of medieval music history. In the fourteenth century there was a profound shift in musical style from the previous century’s *ars antiqua* (the “old art”) to what was termed the *ars nova* (the “new art”). The *Ars nova* was the medieval “avant-garde” with a sound that combined new rhythms, harmonies and texts in complex structural and formal layers. This complexity was due in large part to the expansion and reformulation of the musical notation system. The *Ars nova* theory treatise was a short technical manual that contained rubrics on how to interpret this new notation system. In the traditional historical narrative, the supposed author of this treatise was the composer and poet, Philippe de Vitry (1291-1361), who wrote music in the new style, and was quickly crowned through the annals of music history as the figurehead and putative creator of the ars nova movement.
This narrative is extremely simplified. There is no “one” complete text of the *Ars nova*, but in fact, a small handful of related, but widely divergent, texts extant in manuscripts dating from the fourteenth and fifteenth centuries. The sixteen or so texts that present these new notational theories vary in many ways: in levels of completeness (many of the texts start or break off mid-treatise), in the order of topics presented, in prose style (for example: discursive vs. bulleted-list), and in the textual content itself, that is, the actual words and phrases used to describe specific concepts. There is in fact little hard evidence that proves that Philippe de Vitry actually penned a treatise called the *Ars nova*, and the extant texts may in fact represent remnants of a fluid teaching tradition that may or may not have originated with Vitry. Editions of many of these texts may be found today in various edited volumes, journal articles dating from 1908, 1929 and 1958, and in the nineteenth-century collection *Scriptores de musica medii aevi* edited by Edmond de Coussemaker. However, the various presentation formats, specific editorial policies and accessibility issues have obfuscated attempts at the analysis and interpretation of these texts.
In this paper, I discuss how the complex web of relationships between these sixteen texts may find its best representation in digital form. I focus my discussion on a digital edition I am preparing of three texts. following TEI guidelines, from the *Ars nova* tradition found in these manuscript sources (US-Cn 54.1; E-Sc 5.5.25; I-Su L.V.30). I plan to include annotations that link these texts to the thirteen other related texts of the *Ars nova*, following the Open Annotation Collaboration Data Model. This edition will also have an impact on the broader field of medieval studies as it offers a small-scale study of how the web of quotations, authorities and allusions in medieval texts could be made more transparent and accessible through the use of these types of digitals tools, such as the annotation model. The current process of discovering relationships between texts relies to a large extent on serendipitous discoveries in the footnotes of scholarly articles. The rate of discovery and the level of analysis of the medieval web of textual allusion would increase exponentially with the increased availability of electronic editions, especially if these editions are overlaid with annotations recording and mapping out relationships between texts. It is hoped that this paper will offer an example of how to present this particular textual tradition and others like it.
Many special characters in medieval manuscripts have created problems for online transcription, including glyphs for contractions, medieval punctuation, bracketing, and other non-alphabetical elements. Unicode characters for these elements exist, if at all, only in private forms such as MUFI. Even relatively common characters such as thorn and yogh can easily turn into empty boxes depending on the browser. Up to now, consequently, scholars hoping to publish more than a simplified and normalized transcription of a manuscript in digital form have relied on CDs, on which they can load custom fonts. Even this strategy, though, does not offer the choice of viewing a transcription with contractions (a pervasive feature in many medieval manuscripts) in the original glyphs or expanded. Furthermore, coding XML documents in TEI for such a choice is quite onerous in manuscripts where scribes use many contractions—as much as doubling the coding for a new online facsimile and transcription of British Library MS Add. 61283, the sole witness to The Book of Margery Kempe. The coding team for this project have uncovered strategies that solve both those problems: automating the XML workflow, and embedding an open-source font.
Solution 1: XML Automation
Among the problems for coding contractions: some occur with high frequency, others are occasional at most. The team developed a method to automate a first pass for coding: 1) using Stéfan Sinclair and Geoffrey Rockwell's Voyeur Tools, the team identifies which words occur with the most frequency in the manuscript; 2) the team compiles, from Voyeur Tools analysis, a custom find-and-replace-with-code catalog; and 3) the team imports this catalog into DigitalVolcano's freeware TextCrawler, which is then used to transform the initial transcription into Oxygen-ready coded form—automating the team's "first pass" and encoding in seconds what would have taken the team weeks, if not months, to do without the software.
Solution 2: Medieval Fonts
Encoding medieval manuscripts for diplomatic transcription in XML is challenged by scribes' use of certain characters (e.g. the thorn [ ], yogh [ ], punctus elevatus [ ], punctus elevatus diagonalis [ ], punctus versus [ ]) for which there may be neither Unicode character nor actual font support.
For scholarly works created for publication in print and/or for distribution via common document types (e.g. DOC, PDF, RTF, WPD) and/or electronic media (e.g. CD, DVD), this problem has largely been resolved by the Medieval Unicode Font Initiative (MUFI)'s character recommendations—the latest of which, Version 3.0, proposes the appropriation or addition of 1,548 characters from/to Unicode for use by medievalists—and through the cooperation of font developers, who have created fonts (e.g. Junicode, Andron Scriptor Web) both supporting MUFI's character recommendations and embeddable within these documents and/or distributable on (and, thereby, accessible from) these electronic media.
For scholarly works created for publication to the Web, however, the problem remains largely unresolved. Medievalists have had either to provide as a download to their sites' visitors the font(s) supporting the medieval characters used in their scholarly work or to find a way of representing these medieval characters graphically (e.g. as GIFs, JPGs, PNGs, SVGs), as opposed to typographically. Downloads require documentation and technical support when they fail to result in the successful installation of the font(s). And graphical representation can be both tedious (e.g. for each character, a second, third, or fourth graphical copy may be required to represent its later addition or deletion or appearance in a different size later in the scholarly work) and require certain compromises on the part of the medievalist (e.g. in terms of how these graphical characters should appear when/if the Web page on which they appear is zoomed and/or when only the text on that page is zoomed).
In wrestling with the problem while encoding The Book of Margery Kempe for scholarly publication to the Web, Southeastern's Kempe Project Team devised a method for embedding MUFI-compatible fonts directly within their project's Web site.
Their method relies upon an understanding of 1) the process(es) through which different Web browsers (e.g. Internet Explorer, Safari, Firefox, Chrome, Opera) can render fonts in a Web page; 2) the types of fonts (e.g. EOT, OTF, SVG, TTF) that can be rendered by these different Web browsers; and 3) how widely different even fonts identified by the same name may be from application to application (e.g. Corel WordPerfect to Microsoft Word to OpenOffice.org Writer) and operating system to operating system (e.g. Linux to Mac to Windows).
In the Clarin project, a flexible metadata scheme has been proposed [2, 1]. MPI proposed an implementation 1 that relies on W3C schemas, but according to us, this implementation lacks flexibility. We thus proposed an alternative ODD-based implementation. Unfortunately, ODD lacks two essential features for this task: cardinalities restrictions and interleaving. In this abstract, we will describe the initial problem and sketch the two extensions we added to ODD.
An important part of the vocabulary of politics in French language comes from medieval translations from Latin and Greek into French during the 14th and 15th centuries. This period favoured neologisms in the political science field because translators faced Latine or Greek concepts which did not exist yet in French . This corpus of medival translations is only partially edited. In particular, City of God, written by Augustine of Hippo and translated and commented by Raoul de Presles, which is a major work in the history of ideas in the West, was never edited until now. In order to study the History of the French Political Science Lexicon (HFPSL, acronym of the project by Erc), our team took in charge the edition of this huge text. The main manuscript belonged to the royal library of Charles V, (BnF: Biblioth`que nationale de France, department of Western Manuscripts, Fr 22912 e
(P1 ) and Fr 22913 (P2 )).
The edition of the 894 folios of our main manuscript will allow researchers to have an access to this text in order to lead new researches in linguistics, history and political sciences, and more generally in Humanities. In the paper, we discribe the characteristics of the edition, the TEI encoding and the tools developed.
TEI is one of the most developed tools for analyzing texts on the micro-level and for data mining a large mass of texts. Yet how is TEI-enhanced software being utilized by humanities scholars for their research? This paper presents analysis on the use of MONK, a text-mining software that utilizes TEI-encoded texts to facilitate quantitative analysis of literary texts. The study examines the research conducted in the database using twelve months of website transaction logs from 2010 and series of interviews with researchers who use MONK in their research.
MONK is a text mining research tool hosted by the University of Illinois at Urbana-Champaign Library that enables humanities scholars to mine data from TEI-A encoded texts in select literary databases and archives. MONK builds upon two previously developed text mining programs NORA (http://www.noraproject.org/) and WordHoard (http://wordhoard.northwestern.edu/) in order to create a powerful new environment that “lets users carry out complex data-mining and query operations across collections that contain nearly 200 million words” (MONK documentation, www.monkproject.org/background.html). The SEASR (http://seasr.org/) environment provides the tools for statistical analyses in MONK.
MONK contains a selection of texts spanning from the sixteenth century through late nineteenth century that are encoded in TEI-Analytics or TEI-A, a TEI markup specially created for analytics, via the Abbot tool. Abbot ingested the TEI source files for the texts and normalized them into TEI-A Researchers can also encode and import texts into MONK with the use of Zotero and a MONK Firefox extension.
DATA AND METHODOLOGY
Statistics about the web traffic and usage of MONK were gathered using AWStats, a web log analyzer used to track the web statistics for MONK. The statistics used in this study were gathered January 2010 through December 2010. The investigator is also conducting interviews and surveys with researchers who use MONK, and the analysis of the qualitative data will be completed this summer.
The statistics analyzed include the number of visits on each webpage within MONK, the amount of data processed through each page, number of entry and exit visit, length of the visits, and users' geographic locations.
The statistical analyses conducted on the data included calculating the mean of users that accessed each section of MONK; the mean amount of data transmitted through each page; the distribution of accessed MONK webpages, which were coded as Orientation, Workset, and Toolset webpages; and the frequency of entry and exit points among these three types of webpages.
The quantitative analysis has revealed that the text mining tools in MONK are primarily being used to compile work sets and conduct preliminary statistical analysis. The most frequent tools used on average were:
https://monk.library.illinois.edu/secure/get/CorpusManager.getWorkList = compiling the worksets
https://monk.library.illinois.edu/secure/get/ProjectManager.getToolSets = selecting a toolset
https://monk.library.illinois.edu/cic/public = the opening page
One hypothesis that might be drawn from this data points is that many users are in the initial exploratory steps of using MONK by creating accounts and putting together their first worksets to analyze. Another point of note is a comparison of the accessed MONK pages, which reveals that the use of analytics toolsets and the use of tools for creating worksets were accessed by researchers at a proportion of 2 to 1. In another analysis, the largest amount of data was utilized for a tool comparing the frequency of word features, with a usage of 272.1 MB on average and 15% of the total data processed. These are only a sample of the analyses conducted so far.
This initial examination has begun to reveal several early insights on how scholars are conducting textual analysis research in MONK, and how TEI-A is a critical component of conducting data mining across mass texts. Ultimately, this study critically reveals new avenues of analyzing the research use of TEI in generating quantitative data for textual analysis, and how TEI can be leveraged even further in digital humanities tools.
This paper presents an extensive and complex approach for the analysis and correction of double-keying errors, which is currently applied by the DFG-funded project “Deutsches Textarchiv” in order to evaluate and increase the correctness of text transcriptions and annotations of historical text. Statistical analyses of the error detection and correction results based on a large amount of analyzed text will be presented in order to verify and specify the common accuracy rates for the double-keying method.
In this paper, we will present DeRiK (‘Deutsches Referenzkorpus zur internetbasierten Kommunikation’), a common initiative at TU Dortmund University (Michael Beißwenger and Angelika Storrer) and the Berlin-Brandenburgische Akademie der Wissenschaften (BBAW; Alexander Geyken and Lothar Lemnitzer). The goal is to produce a module of sufficient size and diversity of Computer-Mediated Communication (CMC) as a complement to the reference corpus of the DWDS project.
Since all resources of the DWDS are encoded in compliance with the TEI standard, we want to use and customize TEI for the appropriate base-level annotation of the CMC sub-corpus, thus allowing users of all types to supply their own information on top of this base encoding.
In our paper, we will address the challenges and problems of integrating the CMC component into the DWDS framework and discuss the potentials and the restrictions of the encoding options provided by the TEI-P5. It is our goal to define an encoding scheme for CMC genres which serves the lexicographical requirements that arise from the work at BBAW as well as the needs of linguistic CMC research that have been outlined in paper 1 in this panel. Suggestions for solving some of the problems discussed here will be presented in paper 3 of the panel.
Markup seems on the downturn—the more comprehensive our digital collections of humanistic artefacts become, the higher the success rate of automated analysis. Instead of pre-categorizing and tagging texts in terms of human defined high-level criteria and taxonomies we are now able to retrieve relevant results from the raw data hic et nunc with the help of a tokenizer and some mathematical wizardry. In many users’ everyday digital practice authoritative, rigid top-down directories have long been replaced by the extremely flexible, multi-variable bottom-up oracles of Google et al. which no longer force us to conceptualize our field of interest along somebody else’s lines—or so it seems. Is this where we’re about to go in DH, and in particular in literary computing? Is TEI just an encyclopaedic rear guard action trying to fight off the stochastic forces? As far as expository (non-aesthetic and non-fictional, domain specific) texts go: may be. As far as literary texts are concerned: certainly not. In this paper, we will
The central concern of the edition of Goethe's Faust (https://faustedition.uni-wuerzburg.de) is the exposition of the work's genesis. In accordance with established editorial practice, a distinction between different levels of degree of interpretation is made: The presentation of the ''record'' enables a reader to follow and judge the editorial ''interpretation''. In our case, the representation of the record comprises a detailed account of material aspects of a manuscript and the topography of its inscription. The representation of the editorial interpretation comprises encoding of textual structure, textual modifications and inter-document genetic relations.
In the process of encoding, it quickly became evident that the structures of the two views on the text are disparate, which is to say in terms of markup, they overlap or do not nest properly. This is a well-known problem in text encoding. It is discussed and several solutions are suggested in the TEI Guidelines (http://www.tei-c.org/release/doc/tei-p5-doc/en/html/NH.html). Employing a workaround technique such as using ''milestone elements'' would have sacrificed many of the benefits of XML, among the most important of which would have been formal validation and human readablility and processability. For this reason, we decided to encode both levels as seperate XML files, which are to be combined during processing automatically. This approach is mentioned as ''Multiple Encodings'' in the Guidelines (http://www.tei-c.org/release/doc/tei-p5-doc/en/html/NH.html#NHME). The challenges are to develop a suitable collation algorithm and implementation which correlates parts of the two files, as well as a suitable intermediate data structure that holds the results and can be queried adequately.
The presentation of the manuscripts provides a basis for the genetic analysis and as such must be carried out with care and in a detailled fashion. First, a high-resolution facsimile of the manuscript will be made accessible. Second, a diplomatic transcript lowers the barrier to reading the 18th- and 19th-century manuscripts. On this basis, a more elaborate editorial reconstruction of the genetic process is provided.
The encoding of the manuscript transcripts follows the TEI Guidelines in combination with a recent proposal of the Special Interest Group on Manuscripts (http://www.tei-c.org/Activities/Council/Working/tcw19.html). It captures aspects of the textual structure as well as the materiality and topographical layout of the manuscript. From this information, a diplomatic transcript is automatically produced. It presents the manuscript not in its immediate material and topographical conditions, but in a deliberately idealised form. The talk will discuss challenges and solutions to encoding and presenting manuscripts with features such as erasures, revision groups, multiple writers, transpositions of written text, graphical marks, different writing directions and overlapping inscription. The first challenge presented by this approach is the encoding of complicated manuscripts in a standardised way; the second challenge is the evaluation of layout constraints to produce a pleasing diplomatic transcript.
While using the TEI dictionary module to encode digitized print dictionaries has become a fairly uncontested and very common standard procedure, using the very same system for NLP purposes is quite another story. Our paper will report on a project creating glossaries and dictionaries which are intended to be usable both for human readers and particular NLP applications. It will comprise two parts: in the first section, the authors will try to answer the question why they use the TEI dictionary module as their preferred means to go about the task, also discussing standards and tools such ISO TC 37’s Data Category Registry (ISOCat).
The second part will attempt to pinpoint some encoding issues arising from the projects under discussion such as e.g. keeping track of production metadata or how to include corpus derived statistical data. Another important issue is internal linking and how to reuse examples at various points in the dictionary. We will try to show how the produced data can be delivered as part of a service oriented lexical information system.
In the world of digital dictionaries a great number of different formats coexist: MULTILEX, GENELEX, OLIF, MILE, LIFT, OWL, ISO 1951, LMF (ISO 24613:2008) and the TEI’s chapter on dictionaries whose authors had a very wide range of different applications in mind:
“... The elements described here may also be useful in the encoding of computational lexica and similar resources intended for use by language-processing software; they may also be used to provide a rich encoding for wordlists, lexica, glossaries, etc. included within other documents.” (TEI P5 p.251)
The ICLTT’s collection of digital textual resources also comprises some smaller dictionaries/glossaries which are mainly of historical interest. Currently, efforts are being made to make this data P5 compliant. However, apart from digitizing paper dictionaries, the department has also started a second line of development creating digital lexical resources which–in part–build on the department’s large digital text collections. A number of monolingual and bilingual glossaries and dictionaries are being prepared which are being compiled for particular specialised purposes. Among these resources, there count a glossary of Austriacisms (by which we understand words or phrases considered typical of the German variety spoken in Austria), and a comprehensive dictionary of modern Persian single word verbs. Both resources are being worked on at the moment and both are designed to be used in two scenarios: (a) serving as source to be queried and read by human users in a web-interface and (b) providing data that can be used in NLP applications. The Glossary of Austriacisms is planned to be utilized by tools performing automatic lexical analyses of the Austrian Academy Corpus, a large digital text collection being maintained by the ICLTT. The above mentioned specialised verb dictionary is supposed to furnish data for the creation of a full-form lexicon which in turn is intended to be applied in a morphological analyzer.
This paper attempts to question the encoding of textual fragments using the TEI guidelines. The fragment corresponds to part or a (small) portion of a whole that is missing. Whether it is a rest of an object that disappeared or an unfinished embryo of a work in progress, the fragment is transmitted to us disconnected from the complete and finished opus that would give him his nature, function, and finality. Since the TEI is an encoding scheme that views the text as an ordered hierarchy of content objets (OHCO), as it has been analyzed thorougly, it's not possible to use it without giving each element a tag situated in that hierarchy, and therefore its semantics and functionality. In that perspective, one can ask wether or not the choice of the TEI as en encoding scheme can be misleading and produce as a result improper interpretations ?
But above all, editing fragments consists in establishing it in a set that will determine the way they are read and interpreted. A new signification will be necessarily induced by the new configuration. Can this presentation bias that promotes groundlessly one order in a textual hierarchy superior to the others be avoided ? The solution may be found in the dynamic edition, the one that can offer every possible presentation without imposing one as more important than the others. The critical electronic edition of the documentary files of Gustave Flaubert's last novel"Bouvard et Pécuchet" relies on that viewpoint. This project aims to propose an edition that could give the fragments of citations he collected and started to organize the mobility they deserve, as the volume was very far from finished when he died.
For this project, the TEI is used very pragmatically with two goals. The first consist in "recording" a base structure corresponding to the way these fragments are scripted on the pages of the manuscripted. The second is to use it as a base for en extraction of the editorial units the edition will present away from the original context of the page.
This strategy was only possible because the abstract models of the inscription of the fragments and of the edition that has to be made were clearly established. It shows that strongly embedded markup which is often depreciated, provided that project the only efficient way of extracting these fragment with oll the contextual information that is necessary for a reader to make sense of them, in an dynamic edition that tries to avoid the presentational bias of the printed edition.
The Register of 17th and 18th century Slovenian manuscripts is a TEI (5) encoded archive of ms. descriptions and related digital facsimiles and is the first specialised digital collection of manuscript material in Slovenia, <http://nl.ijs.si:8080/fedora/get/nrss:nrss/VIEW/>.
In the paper, we give an outline of the content and structure of the register, and comment on some specific TEI encoding practices, used in the construction. The archive is focused on early-modern mss., esp. from the periods of baroque and enlightenment. In these periods, a great variety of textual genres and literary forms developed, where texts emerged in several, most distinct socio-cultural contexts, reaching from writings of civil and ecclesiastic persons to texts of peasants and self-educated writers. In our analyses, expressed through precisely structured ms. descriptions, we wanted to capture as many of these aspects as possible: not only codicological and historical data, but also formalised, searchable expressions for literary genres and socio-cultural background of particular ms. To this end, we prepared a taxonomy for textual genres and one for socio-cultural contexts, and linked the ms. descriptions to them. In this way, the msDesc element is not only a container for structured codicological description, but also a carrier of textual and socio-cultural analytical information.
Another aspect of the archive is the gradual preparation of transcriptions of selected manuscripts. In the process of editing, two transcriptions are prepared, a diplomatic one, and a separate critical (edited) text. Besides regular text-critical features, a problem specific for mss. appeared: some portions of text in the source are mixed up and subsequently marked by the scribe in due order. In passages of this kind, the order of diplomatic and critical text differ substantially – but the encoding still has to allow for a parallel presentation of the transcriptions.
The descriptive, analytical and editorial perspectives to the manuscripts shed new light to each other in the form of a hermeneutic circle. We tried to grasp this semantics in the TEI encoding to unfold the complexities of the baroque and enlightenment manuscripts: unpublished and nearly unknown, but rich with research potential and aesthetic value.
TXSTEP offers an interactive XML-based interface to the proven and powerful routines of TUSTEP, the Tübingen System of Text Processing programs. For more than 35 years TUSTEP is being developed and maintained at Tübingen University's computing centre. TUSTEP is a scripting language
as well as a publishing system for the humanities, up until today unmatched in it's overall performance and flexibility. TUSTEP primarily addresses users in the fields of the textprocessing humanities, such as computerlinguists, -philologists and editors. For more information, see www.tustep.org.
But, since it's genuine syntax is proprietary, not intuitive and supposed tobe difficult to learn, users tend to help themselves with other - often less effective - tools or less specific programming languages.
TXSTEP now gives a good answer to this situation by providing a user-friendly XML-syntax, allowing beginners and advanced programmers to use the whole scope of TUSTEP services in a modern, established programmers environment. The benefits are obvious: support of an open standard, widespread dissemination, programming in every other XML-editor, syntax highlighting, code completion and intelligible APIs. Moreover, TXSTEP is aided by the fact that there is no need to change the program's actual core. TUSTEP itself is open source as TXSTEP is going to be as well.
The purpose of TXSTEP, as well as of TUSTEP, is not to provide ready-made solutions for pre-defined problems. It "only" provides program modules for the basic functions of text analysis and processing.
It is the user who has to combine them in order to obtain the solution to a problem at hand. This is the prerequisite that he can take over the responsibility for every detail of the results obtained by computer application.
One of the features of TXSTEP is it's capability to process almost all forms of textual data, whether this being XML-data or plain text files. Wherever there is textual data that has to be processed in the first place in order to gain TEI-data or to enhance the markup of insufficiently tagged XML data, TXSTEP is at it's place.
The proposed demo is based on a prototype and shows the achieved state of our work in progress. It will demonstrate TXSTEPs functionality on the basis of tasks which can not easily be performed by existing XML tools, including problems presented recently on the TEI list.
In the past years two complementary but somewhat diverging tendencies have dominated the field of digital philology: the creation of models for analysis and encoding, such as the TEI, and the creation of tools or software to support the creation of digital editions for editing, publishing or both (Robinson 2005, Bozzi 2006).
These two tendencies are not necessarily mutually exclusive, as the creation of models can represent either the underlying structure or an exporting format for the development of tools. However, these two approaches have often been perceived in opposition, as a dichotomy. On the one hand we have the XML enthusiasts, the editors-as-encoders who apply XML markup to their texts and perhaps also develop publication strategies; on the other hand we have those who support out-of-the-box tools (the ‘magic’ or ‘black’ boxes), who proactively seek the development of fully comprehensive tools that present user-friendly interfaces with the explicit purpose of ‘covering the wires’, in particular hiding the much-abhorred angled brackets. But what are the implications of these positions with respect to the future development of digital (or computational) philology? How realistic is it to ask ‘traditional’ textual editors to turn into encoders? Conversely, how realistic and sustainable is the creation of ‘magic boxes’?
In the past I have studied the difficulties and theoretical implications of using a TEI-based editorial model for an editorial team that was highly geographically dispersed (Pierazzo 2010, but presented as a paper in 2008). On that occasion I argued that the development of ‘magic boxes’ is a very ambitious item to have on the digital philology agenda because every edition, every scholar needs a very specialized, tailored set of tools. In the same article I expressed the opinion that, even if the scholars do not feel comfortable in using tags-on-view XML and the TEI, this was the only reasonable approach for digital scholarly editions. A couple of year later, my judgment has been mitigated somewhat. This was brought about largely by the interesting article by Tim McLoughlin (2010, to be read in combination with Rehbein 2010) which presents in an insightful way the difficulties and resistances in turning a consolidated editorial model into a digital TEI-based one, combined with the experience I gained on some collaborative research projects at King’s College London’s Department of Digital Humanities: these together have triggered questions about the role of technology when it comes to digital scholarly editing. As a matter of fact, the evolution of the editor into an editor-encoder has yet to be investigated in full; at the moment it seems that the attention has been mostly devoted to the steep learning curve necessary to master the techniques of encoding in XML but without reflecting on the deep and sometimes unwelcome changes in the editorial work and workload once a new editorial model is undertaken, particularly when that model is based on TEI. This model sometimes sees the editor-as-encoder evolving also in the editor-as-programmer, the editor-as-web-designer and editor-as-(self-)publisher (Sutherland and Pierazzo 2011). These changes in the editorial work and role of the editors necessarily result in somewhat parallel changes in the final editorial products.
On the other hand the claim for the magic box seem to have receded somewhat, and we have witnessed the appearance of the interesting experience of creating configurable and standard-based tools that have the less ambitious goal of trying to help particular stages of the editorial work (collation, creation of stemmas and critical apparatus, transcription, annotation); this evolution is represented at best, in my opinion, by the tools developed within the Interedition (in particular with CollateX) and TextGrid projects.
This paper will briefly present the background outlined above, and then turn to fundamental issues that arise from it about the nature of editors and editing for digital editions. In particular, it will address the following questions:
Failing to understand how technology can really contribute to the editorial work will have serious consequences in the development and ultimately existence of digital editions.
The paper will address these theoretical and methodological questions making use of concrete examples, particularly from the Jane Austen Digital Edition and from the ongoing editorial experience of the Early English Laws project.
Reference, of course, is a crucial element of scholarly editions. It is the basis not only for classical external citation but also for editing itself: identifying, comparing, manipulating and annotating chunks of text – and also for programs that support or automate such tasks.
Markup has gone a long way from its beginnings to its nowaday use. Initially it was intended to encode THE logical structure of a document regardless of its later graphical form. Today, especially in editions, the markup has the task of preserving a multitude of interests for annotation, e.g. philological, linguistic, historical ones. Fundamental is the fact that with different and maybe evolving or changing interests for annotation the idea of a static, unchanging document, which was the concept of generalized markup at first, is vanishing. It is even questionable if it makes sense to conceive it as ONE document.
In this context an XPath-expression is no longer a sustainable reference. It may be changed by each later inserted tag. Especially programs for the retrieval or transformation make heavy use of the XPath. Realistically the text nowadays has to be seen as a work in progress. And in the present conception this work has the tendency to spoil the fundaments of its own reference system.
In my opinion a solution consists of two parts:
One is to refer only to relatively stable parts of the structure of a document, essential meaningful elements, leaving aside the tagging of annotations etc. In many texts this elements could be paragraphs, sentences and words. These, or appropriate similar ones for special text species, should be the base as well for external citation as for the internal, manual or programmed, editing of the text.
The other thing is considering a radically new organisation of annotation. A fundamental different approach to representing annotations as inline markup is referred to as the “stand-off” annotation model. In a “stand-off” annotation model, annotations are represented as objects of a domain model that “point into” elements of the unstructured content) rather than as inserted tags that affect and/or are constrained by the original form of the content.[UIMA]
There are undeniable advantages, especially in the combination with the suggested basic markup of meaningful elements. Firstly these may serve as elements the stand-off markup can refer to. The next strong point is that annotation can be handled and searched independently of the central document. Furthermore, within the stand-off markup there is no need for overlapping and finally the combination of a linguistically organized basic structure and separate supplementary metadata is a perfect basis for a semantic approach to texts.
This has not gone unnoticed. TEI acknowledges the role of stand-off markup and also other authors, including Robinson, Sahle and myself, are mentioning a potential need for stand-off markup in recent papers. Most remarkably there exists already a standard for this sort of annotation, the Unstructured Information Management Architecture (UIMA). Its Common Analysis Structure (CAS) consists of two fundamental types of objects
Wherever possible UIMA is following established standards. It formulates no domain specific models and so could be complemented by conventions e.g. of TEI.
The potential of this approach has been shown in a popular, nonetheless impressive example by IBM. Its Deep Question Answering System (called “Watson”) is based on UIMA and was able to beat in a prominent US Quiz-Show (Jeopardy!) two human champions. Aside of UIMA DeepQA is using a system of shallow semantic parsing of natural language documents. While UIMA is implemented by Apache mainly in Java (but open for other programming languages), the parsing and also the handling of the database in Watson is done in Prolog. UIMA documents can be transformed in inline markup and/or in RDF. This can also be done selectively, producing individual representations or views of the original document together with its metadata.
In one respect the implementation of UIMA differs from the idea suggested in this paper. In UIMA an annotation points simply to some character string as region of reference in the original document, while in the approach of this paper the metadata point to a list of symbolic elements (words or even sentences). From a theoretical point of view this seems preferable. If this is true also in practice, only practice can show. Therefore it is a pleasure that we could start an implementation of these ideas within the digital part of the edition of the work of Jean Paul in Würzburg.
It is a substantial challenge of digital text curation that similar but distinct collections sometimes must be made to interoperate by a lossless conversion into a common format such as TEI. While it can be relatively easy to verify losslessness for a small text collection, it becomes more difficult with more texts. Curation and verification routines that rely on individual human scrutiny will not operate at a large scale or in a reasonable amount of time.
In early 2007, the MONK Project began to develop a procedure for batch-converting varying collections of XML-encoded texts into a specialized application of TEI P5 that we called TEI-Analytics (TEI-A). The effort to develop a conversion procedure yielded a command-line application, which was called Abbot. Abbot works by analyzing the XML schema that describes the document structure to which the target collection should be converted. Abbot then uses that analysis--an enumeration of allowable elements and their associated attributes--to programmatically generate an XSLT stylesheet that is used for the conversion.
By mid-2009, Abbot had successfully converted 2,585 texts, all of which were valid according to the TEI-A schema. It is a well-known fact, however, that a text collection in possession of document validity may still be in want of some additional form of scrutiny, if only to verify that no words were inadvertently lost or rendered out of sequence. For MONK and its texts, this scrutiny was undertaken on a selective basis by members of the project team. This approach worked for the MONK texts. In the case of the roughly 30,000 texts produced by the Text Creation Partnership--a collection more than ten times larger than the MONK corpus--the problem of validating markup and verifying textual fidelity becomes clear and new procedures are needed. Distant reading, according to Moretti, is the sort of reading that one does when there are too many texts to read closely. Similarly, "distant verification" becomes necessary when there are too many text alterations, or too many texts, to verify closely and individually.
As a developer (with Stephen Ramsay and Martin Mueller) of the original Abbot software, I have extended Abbot to be able to verify the fidelity of all transformations by measuring the inputs and outputs, and calculating and logging every difference. For each XML node, a log entry is made that records any changes to the node, including: (1) the node name, (2) the names of child nodes, (3) the attribute names and values, (4) the text nodes that are children of the current node, and (5) counts of each of the above. All changes are, by default, made as part of the Abbot transformation pipeline and logged in a file that is produced in comma-separated-values (CSV) format. While a command-line diff operation could potentially be used to perform the task of comparing XML files to their source texts, Abbot adds this functionality as a first-class operation to the processing pipeline. The CSV format makes it a trivial task for a spreadsheet program to calculate the consequence of a given conversion, or all conversions.
Every substantive change to the XML structure or to the text content is recorded. Abbot’s measurement of nodal difference is not based on simple string comparison, which would report trivial differences such as <foo n="1" id="a"/> and <foo id="a" n="1" />. In this example, the order of the attributes is reversed, but the two nodes are otherwise the same. XML differencing applications do exist, but are not sufficient, because they are not able to refer to the functions or templates responsible for a given change. The same pipeline that alters the XML input nodes and writes the output nodes should be able--as Abbot now is--to log all differences.
For Abbot, logging is enabled in each XSLT template and each template is reflective. Templates are created at runtime based on input that is compiled at runtime. They vary depending on the source texts and on the desired output schema (TEI-A, TEI-All, or something else). Reflection in this context produces templates that are advantageous in several ways: they are self-identifying, self-describing, and self-differencing. The log file records the location of every change, a description of the alteration, a quantitative measurement, and the unique ID of each template responsible for the adjustment(s).
The change-logging extension of Abbot, by making the integrity of texts verifiable across transformations, solves an important obstacle to keeping curated data meaningful. Gold asserts that a "great challenge of data curation is ensuring that data, once preserved, remains meaningful either within the same research area or ideally across areas or even across domains." When the happy day arrives, perhaps soon, that we have at our disposal the "million[s of] books" that Crane writes about, we will curate them with precision and care and caution and a complete accounting of alterations.
It has been a target of the TEI Ontology SIG for some time to work out mappings between TEI elements and RDF vocabularies. Most notably, there has been a concerted effort to align the TEI with ISO 21127:2006, the CIDOC conceptual reference model. The aim of this paper is to review how far this can be implemented in practice, with the aim of taking an arbitrary TEI document and extracting useful RDF assertions in CIDOC CRM, and thus enabling TEI digital texts to participate more fully in the world of open linked data.
There are three aspects to this work:
We may note that this work is in contrast to linguistic analysis of text, and extraction of assertions from published literature using NLP.
The background to the current effort is the CLAROS project at Oxford which aims to combine discrete
databases of information about the ancient world using an RDF triplestore of assertions using CIDOC CRM. It includes art objects, archaeological sites, antiquarian photographs, and onomastics, and the latter data comes from the "Lexicon of Greek Personal Names" which is available as TEI XML. Getting LGPN include CLAROS involved setting up a TEI to CRM workflow.
A basic tool for recording mapping is provided by the TEI in the form of the <equiv> element, which allows us to provide a specification in ODD which points from a TEI element to an external identifier, and says how to get there. This allows us to build a sensible extraction tool. There are many ambiguities and details to resolve along the way, especially when we work with less structured, but richly marked-up, text.
Among the problems one encounters are
spatial coordinates and dates, the precise expression of which are left vague in CIDOC CRM. The paper gives examples of the results obtained from a variety of TEI texts, demonstrates the implementation in OxGarage of a useable converter, and shows how the resulting RDF can be queried.
Annotating literary texts based on the recommendations of the TEI  can become very cumbersome when editing XML files directly. The double-end-point attachment method, e.g., requires in-depth knowledge of XML and introduces a lot of markup. The basic task of creating an annotation however is very similar to formatting text in word processing programs, where it is convenient to just select some text with the mouse in order to change its appearance or take further action on it. In this paper we present a web-based tool for rapid annotation of TEI documents by just selecting passages of text in a purpose-made preview.
Crafting the markup of passages of text together with links to comments and annotations is a tedious and error-prone process when done directly in XML. Even sophisticated XML editors like oXygen  with support for unique XML identifiers, tag and attribute auto-completion fail in making this procedure a more pleasant experience. One of the proposed methods of the TEI guidelines – the double-end-point attachment method – requires two XML tags (attributed with unique identifiers) to indicate the start and end of a lemma. The annotation itself – contained in another tag – is then linked to the lemma using these identifiers. The built-in annotation method of our system completely hides this rather complex workflow from the user by providing several tools that mimic the behavior of annotating a printed text in reality using pens and labels in different colors and sizes. In doing so it is very similar to the above mentioned, comfortable, and widely-used method of formatting text in word processing programs: the user selects a phrase with the mouse pointer and then takes further action on it like boldening or coloring it differently. Provided that we have an adequate preview of the XML, the user just selects a text snippet in the preview and the annotation is created automatically in the background using the sophisticated methods proposed by the TEI guidelines. Similar approaches are well-known with respect to PDF documents and web sites [3,4].
Following this idea we built a proof-of-concept application for rapid annotation of TEI documents. This web-based application provides easy-to-use tools for searching, highlighting and commenting of passages of TEI encoded text. Thus it is applicable to researchers who might have little or no knowledge of XML and the TEI guidelines. Given a TEI document to be annotated we provide a suitable preview. While exploring the preview or using our infix search we offer virtual felt-tip pens and adhesive labels for annotation. Beside the standard highlighter the user can create additional ones with customized names and colors. All annotations are listed in tabular form next to the preview and are connected with the corresponding positions in the preview through linking. Behind the scenes each modification by the user is sent to the server where the XML processing occurs. At any time the user can export the annotations themselves, a preview of the underlying document with or without the annotations, or the XML file enriched with TEI conformant annotations. At no time the user has to manually insert the required TEI tags and their attributes nor edit raw XML code by hand.
We have evaluated this approach on prose and performance texts. Enclosed please find a screenshot of this proof of concept.
The communication will consist in the presentation of The Descartes Corpus Project which aims at an online edition of all the works and correspondence of Descartes. It is led by the team "Identité et Subjectivité" from the University of Caen (Basse-Normandie), in scientific collaboration with the "Centro interdipartimentale di studi su Descartes e il Seicento" from the University del Salento (Lecce), the Centre d'études cartésiennes" (from Paris IV) and the GREYC (Caen) ; and in editorial collaboration with the Presses universitaires de Caen and the Bibliothèque nationale de France (BNF) .
This project, under development, has these main objectives :
The communication will take place in these topics of reflexion :
Collaboratively building a comprehensive library of linked-up Greek classical texts has now become part of our digital horizon. Any such prospect depends on versatile and well- accepted standards. Thus, The First Thousand Years of Greek is a project which aims to provide TEI-compliant, morphologically-tagged versions of most Greek texts from Homer to the imperial age . To be truly naturalised in scholarly practices, such libraries will have to integrate the critical dimension which major digital collections like the Thesaurus Linguae Graecae have set aside so far. Interestingly, the latest release of XML texts in the Perseus Project was presented as a kind of potlatch: classicists should now take up the challenge of using and improving this offering of “Open Content Greek” . Athenaeus’ compilation being one of these texts, our reflections may be seen as a response to this important initiative.
What should be a new edition designed with an awareness of this transforming landscape? Although Digital Athenaeus, our nascent project, is connected to a traditional philological undertaking, we engage in a natively digital editing process. In our view, this material both demands and rewards new approaches. Indeed, we think of this editio princeps electronica as an experiment at the interface between breadth (large-scale, extensive work) and depth (deep encoding, intensive philological work) [cf. 3,4].
In this paper we do not intend to present a complete rationale. Instead, we want to discuss the suitability of TEI encoding in this classical case. In the first part of the paper we deal with the implications of using text mark-up to represent its structural features and to devise a critical apparatus. In the second part we consider those aspects of an edition for which the combination of TEI with conceptual models can be a viable solution, in particular the alignment of multiple text versions and the formalisation of their relationships.
The Deipnosophists, or Learned Banqueters, is first and foremost an erudite digest: Athenaeus, a Greek author from Egypt who was active in Rome around 200 CE, wrote a gigantic miscellany of texts pertaining to the alimentary and cultural components of the symposium. In the course of some 1,500 pages and 300,000 words in the reference Teubner edition, he adduces thousands of quotations, sometimes famous and preserved in fuller versions, often utterly obscure and otherwise lost. The opacity created by this superabundant matter contributed to Athenaeus’ status, that of an ancillary witness which one frequents somewhat resentfully.
In their loosely thematic structure, the fifteen books follow the parts and practices of a banquet. They combine several layers of dialogue and narration, since Athenaeus reports and comments on the meetings of more than twenty characters. At every level, they extend a nexus of quotations from over 1,200 authors. The establishment of the text would be rather simple — with one 9th-10th c., almost complete manuscript of the unabridged version and apographs —, were it not for the existence of a 12th c. epitome whose textual origin remains unresolved. Given its importance as a source for classicists and historians, the work has had several editions, translations and commentaries; due to its daunting bulk and unusual subjects, this amount of scholarly material remains manageable.
In the long run, the aim of our digital project will be to offer the whole Athenaeus dossier in the form of an evolutive virtual research environment. As the editors will be among its first users, it might also be termed a virtual editing environment.
It is essential to make visible the structure by carefully encoding three overlapping structures whose distribution has not been adequately studied: the characters responsible for the speeches and quotations, the topics treated, the authors and works quoted (quotation, paraphrase and allusion should be distinguished, as well as references to extant and lost works). The bearing of context on the interpretation of fragments makes it a requisite. Furthermore, Athenaeus’ reflexivity is largely embedded in the organisation of internal references, comments and sequences of quotations. To enable a valid perspective, we need accurate reading tools.
The elaboration of a critical apparatus raises interesting issues. How to record differences in the frame of a full-blown environment proposing several editions? This is not the same problem as constructing a digital surrogate for printed apparatuses (on which process  can be consulted). Even if we only contemplate the preparation of a new eclectic text, the full collation of the witnesses shall give a clearer view of this tradition: automatic comparisons and statistical data should contribute to the act of editing. They may take us beyond the prejudice against or in favour of Byzantine capacities for conjectures, which is still predominant in the debates on the relationships between the unabridged and the epitomised version.
Thus, we are aware of the potential of handling the “many texts” (incidentally, unrestricted databases of variants are also databases of revealing manuscript mistakes and meaningful erroneous conjectures). Nevertheless, we think that it is crucial to provide the “one text” which many readers want to use as a reference. And its trustworthiness is increased by the adoption of the “Open Source Critical Editions” model, which implies a reassessment of what a critical edition is, in relation to our current capabilities in terms of accessibility, transparency and explicitness. (On the “one text” and “many texts”, see ; on the OSCE model, see ).
Hence, the understanding of Athenaeus’ work vitally depends on contextualisation and is inherently a matter of structure, while the issue of transcription challenges more generally the received notion of critical method. This leads us to read differently a statement by John Lavagnino : “Two of [the requirements of a TEI approach] can be especially problematic: first, you need to understand your texts; second, you need to believe in the integrity and utility of selective transcription.” While Lavagnino referred to stages in a project or cases when TEI should not be used (e.g. when Encoded Archival Description or DocBook formats are more adequate), for other aspects of the project we might want to combine a TEI schema with other frameworks.
What are the other technical solutions from which a TEI-based “virtual editing environment” could benefit? There have been some recent and convincing attempts both to mix a TEI layer with an ontological approach  and to align TEI elements with classes and properties defined in ontologies such as the Conceptual Reference Model of the International Committee for Documentation (CIDOC-CRM) .
The main point we want to make in support of this method is that building a collection of digital texts upon a well-defined conceptual model can prevent us from creating misleading representations, as it was argued by [11,12]. A digital collection which contains both texts entirely preserved in manuscripts and literary fragments preserved only as embedded quotations, for example, will inevitably contain duplicates unless the underlying model defines and implements the concept of citation. Duplicate records will otherwise alter the results of any quantitative analysis, such as word frequency. Given the number of quotations contained in The Deipnosophists, this requirement becomes pivotal for us: we want to be able
to isolate the quotations from the rest of the text, whether for quantitative or qualitative analyses.
Within the architecture of our project TEI encoding is to be combined with the use of the Canonical Text Services (CTS) protocol, on the one hand, and with CIDOC-CRM and Functional Requirements for Bibliographic Record Object Oriented (FRBRoo), on the other. The reason for this choice is also to separate the mark-up of the text, and of each of its “versions”, from the definition of the relationships that exist between those different versions. By versions of the text we mean, for example, the diplomatic transcriptions of the extant manuscripts, or the available critical editions. The CTS protocol is specifically used to align with each other these versions, whereas CIDOC-CRM and FRBRoo allow us to define their relationships (X is an edition of Y, Z is a translation of X, K summarises Y, and so forth).
Finally, the use of an ontology-based layer aims to make more explicit and machine- understandable the statements usually implicit in critical editions. For instance, through a diagram like the stemma codicum, the editor intends to formulate a hypothesis about the history of the text considered. It should be noted that there have been interesting attempts at expressing and encoding such statements by using an event-based ontology like CIDOC- CRM. Such ways of augmenting the TEI might be instrumental in bringing digital editions to their full potential. Really digital and fully critical editions remain a desideratum in the classical field.
 Center for Hellenic Studies. The Free First Thousand Years of Greek, a project directed by Neel Smith. Harvard University, 2008-. chs75.chs.harvard.edu/projects/diginc/first1kyears.
 Crane, Gregory. “Plutarch, Athenaeus, Elegy and Iambus, the Greek Anthology, Lucian and the Scaife Digital Library – 1.6 million words of Open Content Greek.” The Stoa Consortium, December 13, 2010. www.stoa.org/archives/1332.
 Crane, Gregory. “Give Us Editors! Re-inventing the Edition and Re-thinking the Humanities.” Paper presented at the conference The Shape of Things to Come, Charlottesville: 2010. shapeofthings.org/papers/ (draft).
 Boschetti, Federico. “Digital Aeschylus. Breadth and Depth Issues in Digital Libraries.” In AT4DL 2009. Workshop on Advanced Technologies for Digital Libraries 2009, 5-8. Trento: 2009. www.unibz.it/en/public/universitypress/publications/all/Documents/9788860460301 .pdf.
 Mastronarde, Donald J. “Towards a New Edition of the Scholia to Euripides.” Paper presented at the American Philological Association Conference, 2008. Available in his Euripides Scholia Online Edition. euripidesscholia.org/EurSchHome.html.
 Robinson, Peter M. “The One Text and the Many Texts.” Literary and Linguistic Computing 15.1 (Special Issue “Making Texts for the Next Century”, 2000): 5-14. llc.oxfordjournals.org/content/15/1/5.abstract.
 Bodard, Gabriel and Juan Garcés. “Open Source Critical Editions: a Rationale.” In Text Editing, Print and the Digital World, edited by Marilyn Deegan and Kathryn Sutherland, 83-98. Farnham: Ashgate, 2009.
 Lavagnino, John. “When Not to Use TEI.” In Electronic Textual Editing, edited by Lou Burnard, Katherine O’Brien O’Keeffe and John Unsworth, 334-338. New York: Modern Language Association of America, 2006. www.tei-c.org/About/Archive_new/ETE/Preview/lavagnino.xml (preview version).
 Ciula, Arianna, Paul Spence and José Miguel Vieira. “Expressing Complex Associations in Medieval Historical Documents: the Henry III Fine Rolls Project.” Literary and Linguistic Computing 23.3 (2008): 311-325. llc.oxfordjournals.org/content/23/3/311.abstract.
 Ore, Christian-Emil and Øyvind Eide. “TEI and Cultural Heritage Ontologies: Exchange of Information?” Literary and Linguistic Computing 24.2 (2009): 161-172. llc.oxfordjournals.org/content/24/2/161.abstract.
 Romanello, Matteo, Monica Berti, Federico Boschetti, Alison Babeu and Gregory Crane. “Rethinking Critical Editions of Fragmentary Texts by Ontologies.” In Proceedings of the 13th International Conference on Electronic Publishing: Rethinking Electronic Publishing: Innovation in Communication Paradigms and Technologies, edited by Susanna Mornati and Turid Hedlund, 155-174. Milano: 2009. conferences.elpub.net/index.php/elpub/elpub2009/paper/view/158/66.
 Romanello, Matteo. The Digital Critical Edition of Fragments: Theoretical Problems and Technical Solution. 2011. eprints.rclis.org/handle/10760/15592 (pre-print).
This paper is concerned with the relation between the TEI and DARIAH (Digital Research Infrastructure for the Arts and Humanities). These two endeavors are not only based on partially overlapping research communities, they are also acting in a shared context of the digital humanities, where they are participating in some of the major trends and issues. Therefore, analyzing the current areas of overlap and the potential for future interaction may prove to be of interest for the development of both DARIAH and the TEI.
In order to provide some context, we start by laying out the general aims and tasks as well as the organizational, disciplinary and thematic structure of DARIAH. This EU-funded, large-scale project is an international consortium “aiming to enhance and support digitally-enabled research across the humanities and arts” (DARIAH Mission Statement). This aim is pursued in four virtual competency centers which are concerned, respectively, with building domain-specific research infrastructures, fostering digital humanities research and education, developing standards and recommendations for research data, and developing strategies in the area of advocacy and outreach for digital humanities.
It is against this background that the relationship between the TEI and DARIAH will be considered, the TEI being viewed not only as a (de facto) standard for textual data, but also as an institution and a community. When looking at TEI and DARIAH in this way, the number of issues at hand is obviously vast. In the main part of the paper, we would like to focus on three issues that seem particularly relevant: standards and metadata, tool development, and community building. For practical reasons, we will focus on examples from the German contribution to DARIAH, but many of the issues concern DARIAH more generally.
Our analysis is careful to consider, at this still relatively early stage of the DARIAH project, the role the TEI is already playing in DARIAH as well as the future perspectives for the TEI in this project. At the same time, we look not only at areas in which the TEI can serve as a model to the DARIAH project, but also and quite specifically at ways in which DARIAH can contribute impulses to the further development and dynamics of the TEI. In this way, we hope to offer some insight into existing areas of common ground, but also to reflect on further areas of possible cooperation, the consideration of which may help foster the development of both TEI and DARIAH.