LIT shines at recent retrieval evaluation campaigns

At two recent campaigns aimed at evaluating search engines on a variety of tasks, the Language & Inference Technology (LIT) group performed very well. Last week, LIT obtained a number three position on the Topic Distillation task for web retrieval at the 2003 edition of the Text REtrieval Conference (TREC), held annually at the National Institute for Standards and Technology (NIST), Washington. In addition, very competitive scores were obtained by LIT in both the Question Answering and Robust track.

Topping off an already very rewarding week, results coming in from the INitiative for the Evaluation of XML Retrieval (INEX), towards the end of last week showed that LIT's submissions were top ranking. More precisely, the LIT submissions for the "content only" task (finding relevant XML elements for a keyword query) ranked 1, 2, and 3 (out of a total of 58 submitted runs). And LIT's submissions for the "content and structure" task (finding relevant elements in XML documents where the elements have to satisfy additional structural constraints) ranked 1, 2, and 5 (out of 38 submitted runs).

To find out more about LIT's activities, browse to http://lit.science.uva.nl/.