Robert Stojnic <rainmansr <at> gmail.com> writes:
Hi Gautham,
I think mining wiktionary is an interesting project. However, about the more practical Lucene part: at some point I tried using wordnet to expand queries however I found that it introduces too many false positives. The most challenging part I think it *context-based* expansion. I.e. a simple synonym-based expansion is of no use because it introduces too many meanings that the user didn't quite have in mind. However, if we could somehow use the words in the query to find a meaning from a set of possible meanings that could be really helpful.
You can look into existing lucene-search source to see how I used wordnet. I think in the end I ended up using it only for very obvious stuff (e.g. 11 = eleven, UK = United Kingdom, etc..).
Cheers, r.
On 06/04/12 01:58, Gautham Shankar wrote:
Hello,
Based on the feedback i received i have updated my proposal page.
https://www.mediawiki.org/wiki/User:Gautham_shankar/Gsoc
There is about 20 Hrs for the deadline and any final feedback would be useful. I have also submitted the proposal at the GSOC page.
Regards, Gautham Shankar _______________________________________________ Wikitech-l mailing list Wikitech-l <at> lists.wikimedia.org https://lists.wikimedia.org/mailman/listinfo/wikitech-l
Hi Robert,
Thank you for your feedback. Like you pointed out, query expansion using the wordnet data directly, reduces the quality of the search.
I found this research paper very interesting. www.sftw.umac.mo/~fstzgg/dexa2005.pdf They have built a TSN (Term Semantic Network) for the given query based on the usage of words in the documents. The expansion words obtained from the wordnet are then filtered out based on the TSN data.
I did not add this detail to my proposal since i thought it deals more with the creation of the wordnet. I would love to implement the TSN concept once the wordnet is complete.
Regards, Gautham Shankar