In this paper, we report our experiments in the TREC 2005 Robust Track. Our focus is to explore the use of a new axiomatic approach to information retrieval. Most existing retrieval models make the assumption that terms are independent of each other. Although such simplifying assumption has facilitated the construction of successful retrieval systems, the assumption is not true; words are related by use, and their similarity of occurrence in documents can reflect underlying semantic relations between terms. Our new method aims at incorporating term dependency relations into the axiomatic retrieval model in a natural way. In this paper, we describe the method and present analysis of our Robust-2005 evaluation results. The results show that the proposed method works equally well as the KL-divergence retrieval model with a mixture model feedback method. The performance can be further improved by using the external resources such as Google.
ASJC Scopus subject areas