Estimating Structural Relevance of XML Elements Through Language Model

Abstract

Language modeling approaches have been extensively used as an effective way of measuring ad-hoc document content relevance. However, in structured information retrieval (SIR) there is to our knowledge no approach which aims at assessing structural relevance using language models. In this paper we present a language model based on document-query structure likelihood. As the effectiveness of language modeling relies on the associated smoothing technique we experimented two of these techniques. Experiments are conducted on the INEX 2010 Datacentric test set and show the interest of our method compared to official participants to the task.