Heuristic smushing

August 3rd, 2005

We’ve played around some with integrating foaf and foaf-like data and it’s obvious that the hard problem is smushing. While inverse functional properties are an important resource than helps, maybe we can develop some ideas to support heuristic smushing. Perhaps we could develop an ontology that allows one to say something like

If two instances of a foaf:Person share the samefoaf: givenname, foaf:family_name and a pim:PhoneNumber or pim:Address then they are the sname individual with 95% certainty.

Maybe this is a good use for an RDF rule language like SWRL. But, I think it important to be able to express the degree of certainty.

It would also be helpful if individuals can offer their own rules. For example, I’m pretty sure that I am the only person in the world named “Tim Finin”. If I had the right vocabulary, I could assert this and others could use the information as they saw fit.

Link, link.