Tokunage laboratory at the Tokyo Institute of Technology (Tokodai) was working to develop a method of query expansion. When a user is looking for texts in a large database, he has to formulate his desire through a request, which is a series of words. A request expansion consists in completing the user's request with other words before sending it to usual search engines.
The method proposed by the laboratory uses the database called Wordnet that was developed from 1985 in Princeton University under the direction of Professor George Miller. This lexical knowledge source has been widely used for NLP applications, but has never produced wonderful results for information retrieval. The request expansion proposed by Professor Tokunaga's team consists in searching for new words similar to the ones entered by the user in his request thanks to three information sources`:
An experiment to test this expansion has been done in combination with Cornell University's SMART search engine. It showed that the combination of these three sources increased the performances of the search engines, and proved that the sources were complementary.