Formulating good queries for prior art search

Abstract

In this paper we describe our participation in CLEF-IP 2009 (prior art search task). This was the first year of the task and we focused on how to build effectively a prior art query from a patent. Basically, we implemented simple strategies to extract terms from some textual fields of the patent documents and gave more weight to title terms. We ran experiments with the well-known BM25 model. Although we paid little attention to language-dependent issues, our performance was usually among the top 3 groups participating in the task. © 2010 Springer-Verlag Berlin Heidelberg.