Propositional Logic Representations for Documents and Queries: A Large-Scale Evaluation

Abstract

Expressive power is a potential source of benefits for Information Retrieval. Indeed, a number of works have been traditionally devoting their efforts to defining models able to manage structured documents. Similarly, many researchers have looked at query formulation and proposed different methods to generate structured queries. Nevertheless few attempts have addressed the combination of both expressive documents and expressive queries and its effects on retrieval performance. This is mostly due to the lack of a coherent and expressive framework in which both documents and queries can be handled in an homogeneous and efficient way. In this work we aim at filling this gap. We test the impact of logical representations for documents and queries under a large-scale evaluation. The experiments show clearly that, under the same conditions, the use of logical representations for both documents and queries leads to significant improvements in retrieval performance. Moreover, the overall performance results make evident that logic-based approaches can be competitive in the field of Information Retrieval.