[tei-council] facsimile - how to do encode multiple facsimile images per page?

Conal Tuohy Conal.Tuohy at vuw.ac.nz
Thu Jul 26 10:13:14 EDT 2007


Our original proposal recognised that an important use case is where there are multiple facsimile images of each page. 

Since these images may have different margins and different resolutions, they will not necessarily share the same coordinate space, so it will not be possible to assign an unambiguous set of coordinates to pieces of text, using <att>coords</att>. 

Some of the images for a given page may even be <soCalled>detail</soCalled> images, showing only a particularly interesting portion of the page at high resolution, so the images for a page may constitute a patchwork. 

So to provide a single coordinate space for each page, in which locations can be expressed, but still being able to use graphics of different resolutions, it's necessary to explicitly define the relationship between the coordinate spaces of those graphics. The simplest way to do this is to relate the coordinate spaces of each graphic to a single common coordinate space, which is defined independently of any of the graphics.

This is the reason why our proposal included a new element called <pg> (hereinafter known as <surface> which is a much nicer name). If you had multiple images per page, you would have to use the <surface> element instead of just adding a url to the <pb> elements. Each <surface> element would isntead be linked with a pointer to the corresponding <pb>, and would contain the <graphic> elements which are facsimiles of that page. The <surface> would provide a coordinate space indendepent of the coordinate spaces of the facsimile images. Each <graphic> would have @coords to show the region of the page covered by the <graphic>, and a @scale factor to adjust for different resolution images, i.e. if image A had 2x the resolution of image B then it would have a scale factor of 0.5, to normalise the 2 coordinate spaces. 

This allows for a paragraph (for instance) to be encoded with @coords which can be mapped easily onto regions of any of the graphical facsimiles of that page, by some simple arithmetic. The calcluation of which part of which graphics correspond with which pieces of text could be done in XSLT with XPath and maybe using the key() function, but would not require the use of recursive templates to calculate the graphical transformations, which I think is very important. It ought to be simple to process this to produce e.g. SVG or HTML+CSS or XSL-FO versions in wich text and image are aligned. 

That brings me finally to comment on Lou's proposal of a <zone> element which can self-nest, allowing for a series of graphical transformations. While this could certainly work, I think it is overly complex. I don't believe that the recursive markup (i.e. zones within zones) provides anything that couldn't be done without it fairly easily. Whereas the downside is it would be considerably more difficult to process. Not impossible by any means, but in XSLT for instance I am sure it would require the use of recursive templates, because there's no telling how far the <zone> elements nest.  

Con



More information about the tei-council mailing list