UTA and SICS at CLEF-IP'09

Abstract

This paper reports experiments performed in the course of the CLEF'09 Intellectual Property track, where our main goal was to study automatic query generation from the patent documents. Two simple word weighting algorithms (modified RATF formula, and tf•idf) for selecting query keys from the patent documents were tested. Also using different parts of the patent documents as sources of query keys was investigated. Our best runs placed relatively well compared to the other CLEF-IP'09 participants' runs. This suggests that tested approaches to the automatic query generation could be useful, and should be developed further. For three topics, the performance of the automatically extracted queries were compared to queries produced by three patent experts to see whether the automatic key word extraction algorithms seem to be able to extract relevant words from the topics. © 2010 Springer-Verlag Berlin Heidelberg.