The Text-Image-Link-Editor: A tool for Linking Facsimiles & Transcriptions and Image Annotations

The Text-Image-Link-Editor: A tool for Linking Facsimiles & Transcriptions and Image Annotations
Al-Hajj, Yahya Ahmed Ali, Worms University of Applied Sciences, Germany, alhajj@fh-worms.de
Küster, Marc Wilhelm, Worms University of Applied Sciences, Germany, kuester@fh-worms.de
Introduction
TextGrid’s Text-Image-Link-Editor (TBLE “stands for for Text-Bild-Link-Editor, German for Text-Image-Link-Editor”) is used to link segments of text with sections on the corresponding image. A typical application is linking of scans of facsimiles with their transcriptions, though these texts can also be created during the linking process, which allows e.g. also for image annotations. The information on the linking between manuscript fragments and the corresponding transcription is itself stored in TEI. TextGrid is as a virtual research environment (VRE) for the humanities disciplines dealing with texts in a wide sense (philologies, epigraphy, linguistics, musicology, art history etc.). The joint research project TextGrid is part of the D-Grid initiative, and is funded by the German Federal Ministry of Education and Research (BMBF) for the period starting June 1, 2009 to May 31, 2012 (reference number: 01UG0901A).

TextGrid consists of two principal building blocks, the grid-based backend TextGridRep that hosts both infrastructure services and the repository layer for access to research data and longterm archiving, and the user-facing TextGridLab. The TextGrid Laboratory (TextGridLab), a single point of entry to the virtual research environment, will provide integrated access to both new and existing tools and services via a user friendly software [TG]. TBLE is a key component of the TextGridLab that has been under continuous development since 2008 and is by now in practical use.

State of the Art
The integration of manuscript scans with their transcription and indeed the critical edition itself is a desideratum of modern editions: “While some people continue to think of electronic texts as exclusive of images, the fact is that digital images of manuscripts are electronic texts, as well. The most compelling scholarly editions of the future will make full use of markup schemes such as XML [...], but not without extensive integration of images” [Kiernan2006]. In this context TextGrid is not the only project that recognized the need for an tool to facilitate this linking of image sections with transcriptions. The Edition Production & Presentation Technology's (EPPT's) [EPPT] tool box for integrating images and text operates in much the same solution space. [Parker2009] proposes the development of a web-based Text-Image Linking Environment (TILE), and for much the same reasons as the TBLE, namely to facilitate “the linking of images and textual information [which] remains a slow and frustrating process for editors and curators”. [TILE] is currently under development.For the current status of the project cf. also its homepage (link) Consulted 2011-03-14 Unlike the Eclipse-based TBLE, TILE is Ajax-based, extending the Ajax XML Encoder.

Similar in objective to TBLE and developed in much the same timeframe is Tapor’s / the University of Victoria’s Image Markup Tool [Holmes2010]. Both independently decided to use formats based on TEI P5 to store linking information, though at this stage unfortunately not the same. Unlike TBLE, the Image Markup Tool is a desktop program only for the MS Windows platform and cannot be integrated into the TextGridLab.

[Cayless2008] reports on experiments to partially automate the linking between manuscripts and their transcriptions. TBLE plans to integrate similar functionality using OCR technology (cf. below).

As required in [Huitfeld2010], TBLE allows, in Peirce’s terminology, multiple “types” to be associated with one “token”, or in other words to associate one section in a manuscript with multiple, possibly contradictory interpretations / transcriptions. Image sections can overlap, so that divergent segmentations are possible.

As an aside, this type of linking is very different from the research field of automated image analysis and image annotation which attempts to automatically establish key metadata for an image, e.g. by identifying the objects or persons shown on a photo.

Functionality
The following use case is a typical example for working with the TBLE: A scholar wants to publish manuscripts as collection of images, which offer a digital representation of the original work, but also wants to publish his take on its correct transcription in view of establishing a critical edition. The solution is to write the content of these hand written documents as text in a human/machine readable format e.g. XML and this text can be divided into logical related segments for example: verses, lines and then these text segments can be easily linked with the corresponding sections on the images using the TBLE.

The TBLE can be used for:

Linking Existing Texts and Digitized Manuscripts
Text and image are opened, then the corresponding components (text segment and selected image section) are marked by pairs and the linkage is confirmed. The results can be saved as a new file (local or in the TextGridRepository), which contains the linking information (image coordinates, text segment identifiers, URIs of the used text and image files). Sections can be rectangles or arbitrary polygons.

The content of the new created file represents the saved information as a TEI document with an embedded [SVG] section (see section 4 “The TEI-Model” for more details). Once a file is saved, double clicking it reloads used images, texts and links to continue editing. Changing or adding new links as well as modifying the linked text is possible at any time.

Instead of starting out with an existing transcription and linking it with the image data, the scholar can also decide to start from scratch with an empty text file. The new text segments can be inserted stepwise or at once.

Any number of different and possibly conflicting transcriptions and segmentations can be linked against one set of digitized manuscripts.

Annotation of Image Sections and Existing Links
The Text-Image-Link-Editor offers many other useful features, that help annotating specific links or image sections. For example:

It's possible to build logical groups of links (e.g. verses, comments, etc.) using the layers-tool.
A text-direction (e.g. left-to-right & top-to-bottom) can be assigned to the links.
The TEI-Model
The output file of the Text-Image-Link-Editor follows the TEI model with embedded svg description elements. The following is a list, which crudely describes the structure of the TEI document:

<teiHeader> this element could be used to save the metadata of the document.
<facsimile> in this element is the svg element embedded, which keeps the topographic descriptions of images and links.
<body> in the body element are the link groups, that contain the link elements. These link elements represent the relationship between the image sections and the corresponding text segments. The relationship between images and texts and links is represented in the following figure:
The TEI output file of the Text-Image-Link-Editor to describe the relationship between images and texts and links [Maynooth2010]

Full Size Image

Structure and Components
The Text-image-Link-Editor is a part of the TextGridLab and is implemented as a group of Eclipse plugins following the [MVC] patternThe Model-View-Controller (MVC) pattern separates the modeling of the domain, the presentation, and the actions based on user input into three separate classes.

This tool consists of a Toolkit and two views in addition to the XML Editor and the generic Navigator:

Image View: shows the images and enables you to select individual image sections to be linked.
Thumb View: is used for navigation. It displays a reduced version of the entire image and the active image detail (which is enlarged in the Image View) which can easily be moved and zoomed.
Toolkit: provides different functions for working on the Image View.
XML Editor: allows you to open or create texts as well as to select individual text elements.
TBLE in action

Full Size Image

Further Enhancement
TBLE is already actively used in a number of projects, but continues to be enhanced, taking into account new user requirements coming up in the field. In particular, we plan to implement a new feature in the Text-Image-Link-Editor to enable an automated segmentation of facsimiles using the [OCRopus] OCR-system, which offers a possibility to partially automate the linkage process.OCRopus (tm) is a state-of-the-art document analysis and OCR system, featuring pluggable layout analysis, pluggable character recognition, statistical natural language modeling, and multi-lingual capabilities. http://code.google.com/p/ocropus/ . Consulted 2011-03-14

References:
Cayless, Hugh 2008 “Experiments in Automated Linking of TEI Transcripts to Manuscript Images, ” TEI Members Meeting, 2008 (link)

Kiernan, Kevin et al. Edition Production & Presentation Technology (EPPT), 2011-03-12 (link)

Holmes, Martin The UvC Image Markup Tool Project. Project homepage, 2010-10-31 (link)

Huitfeldt, Claus; Yves Marcoux and C. M. Sperberg-McQueen 2010 “Extension of the type/token distinction to document structure, ” Proceedings of Balisage: The Markup Conference 2010, Balisage Series on Markup Technologies 5 doi:10.4242/BalisageVol5.Huitfeldt01.

[Kiernan2006] Kiernan, Kevin 2006 “Electronic Textual Editing: Digital Facsimiles in Editing, ” Electronic Textual Editing, Burnard, Lou; O'Brien O'Keeffe, Katherine and Unsworth, John (Eds) MLA 2011-03-14 (link)

[Maynooth2010] Maynooth - Michael Leuk, Dr. Simon Rettelbach 2010 Cost Workshop,

[Parker2009] Porter, Dorothy Carr; Reside, Duke andWalsh, John 2009 “Text-Image Linking Environment (TILE), ” Digital Humanities, 2009, 2009 p. 388ff

[SVG] W3C, Scalable Vector Graphics (SVG), 2011-03-14 (link)

[TEI] The Text Encoding Initiative, 2011-03-14 (link)

[TG] About TextGrid, 2011-03-14 (link)