Ivan's Blog

Read through the paper of Bechhofer & al[1] from ISWC. Interesting stuff. Use Semantic Web to help accessibility problems. Per se, this is not new, but the approach of Bechhofer & al. seems promising. What they do is (1) extract CSS class names from an HTML page (not necessarily XHTML…) and (2) use ontologies to categorize the role of each element (as referred to by CSS). Using the knowledge gained from the ontology a tool can rearrange the page, remove unnecessary elements, etc. The ontologies consist of two parts: on the one hand a general ontology describing the possible roles of various elements in general ( e.g., menus) and, on the other hand, specific ontologies for a specific classes of pages. Eg, ontologies are available for all the CNN.com pages, or for the blogger.com blogs. (There is a much better description on the project’s home page.)

Having some contacts with the Web accessibility community (eg, via my contact with the Accessibilty.nl group here in the Netherlands) I know how really difficult these things are. Also, the fact that the Web designers’ community has not really adopted XHTML over HTML makes the development of various tools even more difficult. The advantage of SADIe seems to be that it annotates the (now ubiquitous) CSS tags rather than relying on the content of the (X)HTML page; also, the ontology can be defined for a specific Web site (or family of sites) independently so that no extra information has to be added to the site itself.

Although the project is still in research, I was curious, so I tried the on-line facility for some sites. I was not very lucky with the CNN home page; there was hardly any difference. But, when I randomly chose a blog on blogger.com, the difference with the site generated by the tool is noticable (e.g., the menus are rearranged):

I did not compare this with the output of other tools (like Opera’s small screen rendering which does similar things based on very different techniques). But it is certainly promising.

It is interesting that Danny Ayers just published a blog on turning CSS sructures into RDF. The current tool of SADIe is a javascript goody analysing the DOM tree of the page, and using the CSS ontology knowledge to massage the DOM tree. However, if the CSS structure can be turned into RDF, this can then be mashed up with the Ontology defined by SADIe; ie, other tools can do nice things with those data, too (for example, specialized queries could extract the meaningful information from the page, stuff like that). Could become even more interesting…

1. “SADIe: Semantic Annotation for Accessibility”, by Sean Bechhofer, Simon Harper, and Darren Lunn