A Novel Method to Predict Query Performance Based on Cluster Score

Predicting query performance has been recently recognized by the information retrieval community as a crucial issue in information retrieval systems. In this paper, we present a novel method for predicting query performance by computing cluster score. For a fixed query, cluster score quantifies and reflects the correlation between retrieved document collections and each query term and the distribution of this correlation simultaneously. Experiments demonstrate that cluster score significantly and consistently correlates with query performance in a variety of TREC test collections. We compare cluster score with the clarity score method which is the state-of-the-art technique for query performance prediction. Our experimental results show that cluster score performs better than, or at least as well as clarity score.