TALP at GeoCLEF 2006: Experiments Using JIRS and Lucene with the ADL Feature Type Thesaurus

This paper describes our experiments in Geographical Information Retrieval (GIR) in the context of our participation in the CLEF 2006 GeoCLEF Monolingual English task. Our system, named TALP-GeoIR, follows a similar architecture of the GeoTALP-IR system presented at GeoCLEF 2005 with some changes in the retrieval modes and the Geographical Knowledge Base (KB). The system has four phases performed sequentially: i) a Keyword Selection algorithm based on a linguistic and geographical analysis of the topics, ii) a geographical retrieval with Lucene, iii) a document retrieval task with the JIRS Passage Retrieval (PR) software, and iv) a Document Ranking phase. A Geographical KB has been built using a set of publicly available geographical gazetteers and the Alexandria Digital Library (ADL) Feature Type Thesaurus. In our experiments we have used JIRS, a state-of-the-art PR system for Question Answering, for the GIR task. We also have experimented with an approach using both JIRS and Lucene. In this approach JIRS was used only for textual document retrieval and Lucene was used to detect the geographically relevant documents. These experiments show that applying only JIRS we obtain better results than combining JIRS and Lucene.