P5 Guidelines, 22.3 (Specific Elements for Graphic Images)
The P5 Guidelines describe four elements for including graphic images in TEI documents:
figure: contains a block containing graphics, illustrations, or figures.
graphic: indicates the location of an inline graphic, illustration, or figure.
binaryObject: provides encoded binary data representing an inline graphic or other object.
figdesc: contains a brief prose description of the appearance or content of a graphic figure, for use when documenting an image without displaying it.
<figure> is the parent element for the others. <figure> may also contain <head> (a header or title) and <p> (commentary or caption, not a description of the image):
Figure is also self-nesting, which allows for description of image fragments, though as TEI is now described those image fragments would have to be separate image files. Would it be useful to instead have a way to note coordinates using an attribute or attributes in <graphic>? (@mets:coords or @svg coordinate attributes; see TEI_to_SVG#Mapping_attributes):
If a graphic contains a large amount of text, and the editor determines that the graphic actually contains that text (the text is not part of the surrounding text), <figure> may contain <text>.
To link thumbnails to larger images, create a <ref> tag that points to the xml:id of the corresponding figure. Include the thumbnail as a graphic in the <ref>
<ref target="#IM1">Click here
<graphic url="fig1th.png"/>
for enlightenment
</ref>
<figure xml:id="IM1">
<graphic url="fig1.jpg"/>
</figure>
Associate images/parts of images with text elements
Use the linking mechanism described in 14 Linking, Segmentation, and Alignment to link together a figure (the example in the guidelines uses an external SVG file, parts of which are identified with xml:ids) and TEI elements (which are likewise identified using xml:ids).
This document, authored by Richard Gartner and Lou Burnard and last updated in 2001, outlines ways to use TEI to represent three types of digital objects:
complete digital transcriptions of the content of manuscript or print originals (possibly including illustrations as well as text)
collections of digitized page images (digital facsimiles) intended for use as surrogates for complete manuscript or print originals
digital objects (digital editions) combining both page facsimiles and transcriptions, possibly also including layers of editorial annotation
The authors attempt to address these areas of concern:
the need to distinguish images of a manuscript or print source from images located within it;
the need to support multiple formats of a single image;
the need to associate metadata at different levels (e.g. collection level, item level);
the need to associate transcription and facsimile in a standard way;
the need to define practices which can be used equally well in both SGML and XML environments;
the desire to avoid special purpose rules which assume nonstandard or ad hoc processing rules.
Case 1: Transcription
Updated to cite chapters in P5:
The content of a transcription should be marked up as a single <TEI> element using the standard TEI elements <text>, <body>, <div>, etc. from the TEI core tag sets.
If the source contains "significant illustrative material", use <figure> (and children) from Guidelines chapter 22 Tables, Formulae, and Graphics to insert illustrations in their proper location in the transcript.
Example of the TEI <figure> element.
<lb n="12"/>feond mid his geferum eallum. Feollon þa ufon
<lb n="13"/>of heofnum þurhlonge swa þreo niht and da
<lb n="14"/>gas, þa englas of heofnum on helle, and heo ealle
<lb n="15"/>forsceop drihten to deoflum.
<figure>
<head>Fall of the Angels</head>
<figDesc>Above, God, cross-nimbed, beardless, and holding a closed book,
accompanied by three angels, turns and gestures toward three angels on
the left, one of whom (Lucifer?) holds a palm frond ...</figDesc>
</figure>
<pb n="17"/>
Case 2: Digital Facsimile
Should be marked as a separate <TEI> element. TEI structural tags may be used, but need not be, especially if differences between textual structure and physical structure cause overlap.
Each distinct image making up the facsimile should be encoded as a <figure> element, arranged in the normal reading sequence of the facsimile. The appropriate milestone element (<pb/> or <cb/>) should be used at the appropriate place in the facsimile.
Example of the TEI <figure> element with reference to manuscript folio number.
<pb n="16"/>
<figure n="16">
<figDesc>[Page 16: 15 lines of text followed by image of "Fall of Angels"
(Ohlgren 16.11)]</figDesc>
</figure>
<pb n="17"/>
<figure n="17">
<figDesc>[Page 17: 15 lines of text followed by image of "Fall of Angels"
(Ohlgren 16.12)]</figDesc>
</figure>
<figure> elements may be self-nested to show that one image logically contains others (in cases where two image fit together to form one larger image).
Example of nesting TEI <figure> elements.
<figure n="16">
<figure>
<figDesc>[Page 16: 15 lines of text (detail of upper part)]</figDesc>
</figure>
<figure>
<figDesc>[Page 16: Image of "Fall of Angels" (Ohlgren 16.11) (detail of lower part)</figDesc>
</figure>
</figure>
3: Techniques for Referencing Images
Recommends declaring the file containing the image as an external entity and then referencing that entity using the "entity" attribute in <figure>. This approach depends on the use of the DTD (no longer practical given the TEI's move towards using the RelaxNG schema), and relies on an attribute that is no longer included in P5.
In TEI P5 there are also two new elements, <graphic> and <binaryObject> which may be used to represent image references or images respectively. These permit the encoding of figures containing multiple images, for example. The <graphic> element has a "url" attribute which enables the association of a <figure> with its corresponding image file. The recommendations in Gartner and Burnard need to be revised substantially in this respect.
Example of the TEI <graphic> element with reference to image file URL
<pb n="16"/>
<figure n="16">
<figDesc>[Page 16: 15 lines of text followed by image of "Fall of Angels"
(Ohlgren 16.11)]</figDesc>
<graphic url="http://image.ox.ac.uk/images/bodleian/msjunius11/16.jpg"/>
</figure>
<pb n="17"/>
<figure n="17">
<figDesc>[Page 17: 15 lines of text followed by image of "Fall of Angels"
(Ohlgren 16.12)]</figDesc>
<graphic url="http://image.ox.ac.uk/images/bodleian/msjunius11/17.jpg"/>
</figure>
<figure> may contain more than one graphic, so it is possible to relate multiple image files to a single physical object:
<pb n="16"/>
<figure n="16">
<figDesc>[Page 16: 15 lines of text followed by image of "Fall of Angels" (Ohlgren
16.11)]</figDesc>
<graphic n="daylight" url="http://image.ox.ac.uk/images/bodleian/msjunius11/16.jpg"/>
<graphic n="uv" url="http://image.ox.ac.uk/images/bodleian/msjunius11/UV/16.jpg"/>
</figure>
<pb n="17"/>
<figure n="17">
<figDesc>[Page 17: 15 lines of text followed by image of "Fall of Angels" (Ohlgren
16.12)]</figDesc>
<graphic n="daylight" url="http://image.ox.ac.uk/images/bodleian/msjunius11/17.jpg"/>
<graphic n="uv" url="http://image.ox.ac.uk/images/bodleian/msjunius11/UV/17.jpg"/>
</figure>
Other options mentioned in the recommendation: Xlink or Xinclude.
4: Aligning transcription and facsimile
Example of linking image file with transcription using @corresp on <pb>.
The TEI Guidelines propose a number of methods for aligning parts of a document (see www.tei-c.org/Guidelines/SA). These may be briefly summarized as follows
use the @corresp attribute to assert that a <pb> (or other reference point) in the transcription corresponds with a <pb> in the facsimile (or the reverse);
alternatively, use a stand-off <link> element to assert the association between the two <pb> elements;
for cross-document linking, an intermediate <xptr> may be used as the target.
Method used by Edition Production & Presentation Technology (EPPT) to link images and text with complex markup
EPPT (www.eppt.org/eppt) is a software platform designed to help editors prepare and display image-based XML-encoded electronic scholarly editions. EPPT uses electronic text, DTDs, and digital facsimiles (JPG, BMP, GIF, TIFF, etc.) to produce highly complex encoding of image-based textual resources. EPPT supports any DTD-based encoding schema (such as the TEI P4 DTDs – with support for P5 XML Schema and RelaxNG currently under development) and enables precise, semi-automatic linking of full images as well as image sections with structural and descriptive metadata. EPPT includes new technologies and extensions (such as support for multiple hierarchies) to provide a powerful user interface for leveraging XML technologies in the production of image-based XML-encoded electronic editions. See a full description of the EPPT.
The Image Catalog is a simple XML file that consists of lists of image files organized by category. There is one category for every set of image files for a given source.
daylight (images taken under regular lighting)
ultraviolet (images taken under ultraviolet lighting)
microfilm (images scanned from microfilm)
Categories are assigned unique IDs.
Within each category, the user associates each file with an ID that is unique within the category. IDs will repeat across categories. Essentially, these IDs associate the files with the same object, which may be represented by multiple images.
Example of the EPT Image Catalog configuration file.
The file IDs from image catalog are then associated with an attribute within the text markup itself. The attribute used is configured by the user; for the Electronic Boethius, we used the attribute n in the element <fol> (folio). The content for each folio is thus associated with the correct image files (from every category):
<fol n="5v">
points to
<daylight><file id="5v"/>
and
<ultraviolet><file id=“5v”/>
and
<microfilm><file id=“5v”/>
Example of EPT Image Catalog links image files with sections of text.
Attribute for Coordinates
A TEI element can be associated with a rectangular region of an image by encoding the co-ordinates of the region's corners as an attribute of that element. The EPT allows the user to assign an attribute (Electronic Boethius uses coords). In whatever elements that attribute is defined, the coordinates, selected by the editor, are automatically entered by the software. The ID assigned to the categories in the Image Catalog are appended to the beginning of the coordinates.
Example of @coords on <figure>.Screenshot of EPPT
<damage coords="d:1058,738,1093,798">
Method used by the UVic Image Markup Tool: Annotate images
load and display a wide variety of different image formats
allow the user to specify arbitrary rectangles on the image, and associate them with annotations
allow such rectangles to overlap if the user wishes
provide mechanisms for bringing overlapped rectangles to the front easily
allow annotations to be grouped into categories, distinguished by name and appearance, to allow different classes of annotation
require no significant knowledge of XML or TEI
allow the use of XML code if the user wishes
save data in an XML file which conforms to a TEI P5-based schema
reload data from its own files
generate a simple "Web view" of the data for display in a browser
come packaged with an installer, Help file, and basic tutorial
be simple for novices to use!
Simplified example from the IMT web:
<TEI>
<teiHeader>
<!-- ...header content... -->
</teiHeader>
<text>
<body>
<!-- ...TEI content... -->
<div type="imtAnnotatedImage">
<svg xmlns="http://www.w3.org/2000/svg">
<!-- ...svg content... -->
<image>
<!-- Image file is linked in here -->
</image>
<rect>
<!-- Annotation area on the image defined here -->
</rect>
</svg>
<div type="imtAnnotationLayer">
<!-- Annotation content goes here -->
</div>
</div>
<!-- ...TEI content... -->
</body>
</text>
</TEI>
Using this system, a user can annotate many sections of a single image. Unlike the EPT system described above, in which the coordinates are stored directly in the element, here the image coordinates are stored in a separate section and linked to the annotations using unique IDs (svg:id in <rect>) to link the areas to the annotation content (n in <div type="imtAnnotationLayer">).
Example of the coordinate system used by the UVic IMT.
<div type="imtAnnotatedImage" xml:id="NB-B-17328">
<svg xmlns="http://www.w3.org/2000/svg">
<image xmlns:xlink="http://www.w3.org/1999/xlink" xlink:href="NB-B-17328.jpg" width="5060" height="6025"/>
<title>NB-B-17328.jpg</title>
<desc>5060 x 6025</desc>
<rect svg:id="NB-B-17328_0" x="4287" y="1412" width="580" height="426" style="3" color="#ff0000"/>
<rect svg:id="NB-B-17328_1" x="4067" y="2019" width="720" height="747" style="3" color="#ff0000"/>
</svg>
<div type="imtAnnotationLayer" n="NB-B-17328_0">
<head>la chandelle allume la découverte.</head>
<p>Il a trouvé la cache.</p>
</div>
<div type="imtAnnotationLayer" n="NB-B-17328_1">
<head>le couvercle de la malle</head>
<p>Il est pris sans vert</p>
</div>
Screenshot from the UVic IMT interface.Browser view of output from the UVic IMT.
While a library may record descriptive metadata regarding a book in its collection, the book will not dissolve into a series of unconnected pages if the library fails to record structural metadata regarding the book's organization, nor will scholars be unable to evaluate the book's worth if the library fails to note that the book was produced using a Ryobi offset press.” (METS: An Overview and Tutorial)
Why METS?
System to organize disparate parts and relate files (and pieces of files):
Areas on images correspond with sections of text
Areas on images relate to one another
Areas of text relate to one another
METS provides sections for defining the logical and/or physical structure of the digital object
File Section (fileSec)
Structural Map (structMap)
File Section
“lists all files containing content which comprise the electronic versions of the digital object.”
Similar to the EPT Image Catalog
Within the FileSec
FileSec consists of Groups of File Locaters
File Group: contains related files
<fileGrp USE="facsimile">
File: assigns the individual files unique identifiers
<file ID="C-Clbiv-69v" MIMETYPE="image/tiff">
File Locater: points to the location of the file, using xlink syntax
“outlines a hierarchical structure for the digital library object, and links the elements of that structure to content files … that pertain to each element.”
Organization may be logical or physical
A single METS file may contain multiple structural maps
Within the StructMap
StructMap consists of organizational divisions, that may nest
The first file pointer references the corresponding image file, while the second one references the corresponding line range (BEGIN and END) in the XML file (these are the values of, for example, @xml:id on <lb/>
ALTO (Analysed Layout and Text Object) is an extension schema for METS, allowing the markup of layout information of digitised texts. It has had a bit of use in newspaper digitisation in the States, in the National Digital Newspaper Project.
In summary, it defines digitised texts as consisting of "page" elements containing "blocks" which have rectangular coordinates assigned to them using attributes HPOS, VPOS, WIDTH, and HEIGHT, and also a ROTATION. Each block element's page is associated with image files with stand-off markup in the normal METS way.
By contrast we need a different way to relate TEI content elements to pages, since the dominant hierarchy in TEI is logical rather than physical (i.e. div rather than page).
Conclusions
Draft Recommendations for TEI Digital Facsimiles
Not robust
No specific image/text linking
Based on TEI P3 (though updated on this page to TEI P5)
Edition Production Technology (EPT)
Designed for textual annotation, but can annotate illustrations as well
Markup schema is not defined - editors can use their own markup schemas (requires @coords)
Complex
Dependent on editing software
Supports most image file formats (jpeg, tiff, gif, bmp)
No visualization outside of the editing environment
UVic Image Annotation Tool
Simple
Annotates via description – no “text encoding” (i.e., cannot include in an electronic transcription)
Based on accepted standards
Markup schema is defined (extended TEI + SVG)
Visualization through SVG-enabled browser
METS
Very complex
Objects separate from indexes
Flexible – can link many textual and image sections (Venetus A – link main text and several layers of annotations)
No visualization (yet) for the more advanced image-text links
