clauwa.info

Just came back from ESWC2010 (the 7th Extended Semantic Web Conference). This year I was there to present my PhD research at the PhD symposium (my slides can be found here), which turned out to be an excellent opportunity to discuss my PhD work and to get valuable feedback!

Today on the flight back I was trying to figure out what were my personal highlights of the week… Almost all talks I heart were very interesting. Specially, I really enjoyed Claudio Giuliano`s talk about Acquiring Thesauri from Wikis by Exploiting Domain Models and Lexical Substitution. He presented their approach to induce thesauri from Wikipedia by exploiting the content and structure of Wikipedia.

The 2 basic building blocks of their approach for detecting paradigmatic relations are semantic domains and lexical substitutability. For a given Wikipedia concept they extract synonym candidates by extracting frequent terms from the wiki page of the concept and from incoming links. Than they calculate the similarity between each term and the concept by using LSA. For the most similar terms (similarity threshold > 0,5) a lexical substitution test is performed in order to find out whether the terms are really synonyms of the concept label. They check whether these terms can be substituted by the concept label by generating sentences (where synonym candidate terms are substituted by the concept label) and by looking for the frequency of the generated sentences in the Web 1T 5-gram corpus. Candidate words are then ranked according to that plausibility score (see there paper for details).

I think this is an interesting approach, because existing thesauri such as WordNet are heavily used for all kind of Web mining tasks but they evolve slowly and mantaining them is time-consuming. On the other hand Wikipedia has a big community and is therefore very dynamic and evolves fast.

Another highlight was from my point of view the AI mashup challenge. I saw many really cool mashups! Alexandre Passant with his music recommendation application dbrec is the well-deserved winner of the AI mashup challenge.

Finally, it was of course also a highlight for me to get one of the best student paper awards at the closing ceremony (Lora’s slides can be found here - I especially like slide 14 - tweets during the dinner). So all in all ESWC2010 was for me a very interesting and inspiring week with many highlights and no #fail