A NeOn week for Watson

Last week, the review meeting of the neon project have taken place in KMi. In addition to partly explaining my complete absence on this blog recently (not that anybody else is actually present anyway), it was also a great opportunity to discuss to lots of people having the interesting attribute to be interested in Watson. So this time, I will not speak about the server... not a word... not even a tag! I prefer to look a bit more forward and to list some of these ongoing, planned, and envisaged collaborations and system integrations that were mentionned, discussed, or even organized last week.

Note to the people of NeOn: In case I forgot something, don't hesitate to tell me (by e-mail, through comments, etc.).

We are discussing the integration of Watson with Oyster2 for quite a long time (I think since November 2006) with Raul from UPM in Madrid, Spain. Oyster is a peer-to-peer system dedicated to ontology sharing and publication. In other terms, if Watson is the Google of the Semantic Web, Oyster is the e-mule...

The basic idea of the integration of Oyster and Watson is quite simple: ontologies discovered by Watson would appear on the Oyster network and ontologies registered on the Oyster network would be discovered by Watson. Another important thing in relation to this is that, for the purpose of Oyster, Raul and some other guys have developed OMV, the ontology metadata vocabulary. OMV is intended to be the standard for describing ontologies on the Semantic Web and rely itself on OWL. Using the Oyster API, Watson will then be able to generate a standard and machine readable description of the ontologies it collects!

Jerome from INRIA Grenoble, France, is developing an alignment server for the purpose of NeOn. Basically, this server would provide a repository of alignments (correspondences between ontologies) and ontology matching methods. There are many different ways in which this alignment server and Watson may benefit from each other: for example, Watson could call the alignment server to know if an alignment exists between two particular ontologies or can use more advanced matching techniques to compute relations between ontologies. On the orher hand, the alignment server could query Watson to find useful ontologies, or to indentify ontologies that may be easily mapped (because of their similarities).

We are planning to add the possibility for user to review ontologies in Watson by using http://revyu.com (developed by Tom from KMi), and to use the reviews for ranking ontologies. We discovered last week that Holger from AIFB in Karlsruhe, Germany, was developing a trust based system that complement perfectly Watson and revyu.com.

The basic idea of this system is to rely on meta-reviews, stating the trust of a particular user concerning a particular review. The nice things are that 1- these meta-reviews can be managed through revyu.com as well, 2- they allow a more personalized ranking since it can depend on the trust a particular user have on a particular set of reviews or on a particular reviewer, and 3- meta-reviews do not necessarily need to be applied on manual reviews: this mechanism can be a way for users to express their trust concerning the quality measures we use for ranking ontologies, and then, to obtain personalized ranking (this is so cool that I cannot wait for it to happen!)

Then, now I would rather go back to work if I want Watson to be ready for that...