The UPV at GeoCLEF 2007

In this work we attempted to determine the relative importance of the geographical and WordNet-extracted terms with respect to the remainder of the query. Our system is based on Lucene and uses LingPipe for Named Entity recognition. Geographical terms are expanded with WordNet holonyms and synonyms and indexed separately. We checked the relative importance of the terms by boosting them with reduction factors (0.75, 0.5 and 0.25). The comparison to the clean system (using only Lucene) shows that it is possible to improve the mean average precision if the importance of geographical terms is equal or less than the half with respect to the content words in the query. We also observed that WordNet holonyms may help in improving the recall but the term expansion is sensible to ambigue place names. As a further work, we will need to implement a toponym disambiguation method in order to reduce the impact of this kind of ambiguity.