ABSTRACT
The basic processes in a text retrieval system are text representation, representation of a user’s information need, and comparison of these two representations. These processes are complementary, and improving the effectiveness of text retrieval will involve improving them all. Retrieval models provide the theoretical frameworks for integrating research in these areas. In this paper, we give an overview of the basic text retrieval models and then describe a recent model that is based on probabilistic inference. This model has been tested successfully in a variety of retrieval environments and can potentially make effective use of complex text representations produced by natural language processing techniques.
