Towards Machine Understandable Diagrams For Accessibility And The Semantic Web

Hi,

I've found some more spare time, and so thought I'd offer my latest thoughts on accessible diagrams.

SVGT 1.2 should enable a user to be provided with sufficient information that they can then extract the semantics that are encoded within a diagram.  However, SVG, at least in my opinion, doesn't currently support machine based semantic translation of diagrams between different physical representations.  This could be considered important for accessibility, as it mitigates situations where the doc author hasn't provided the content in alternative forms using the title and desc attributes.

At present, SVG does support a limited amount of semantic extraction to be performed on a diagram.  You can run database comparisons between mathematical descriptions of the characteristics of shapes and their semantic meaning, you can perform database look ups for the semantic meaning encoded within spatial relationships.  These are possible as the spatial ontology, which is the set of mappings between concept and physical representation, remain largely static throughout document type, rather than just document.  There are document specific ontology's, such as color, font, font attributes, and relative size that need to be specified on a document by document basis in order to achieve accurate semantic translation.  Again, this could be done using a database look up, although this would be an inefficient process, requiring that a user enter the ontology for each document, or at least each document author.  However, there is an alternative, and that is to encode these document specific ontology's as meta data within the SVG format, although this could lead to doc authors not specifying the meta data.

At the minimum, document type, e.g. UML class diagram or organisation chart, needs to be specified in meta data.  The user could be allowed to specify it, but this would require the user to first establish the diagram type through the use of assistance, and if the user were to specify the doc type then autonomy has gone from the system.  The doc type is crucial to the system working, as it is needed in order to provide the context for the database look up driving the semantic translation process.

Will