Clustering search engine query log containing noisy clickthroughs

Query clustering is a technique for discovering similar queries on a search engine. In this paper, we present a query clustering method based on the agglomerative clustering algorithm. We first present an overview of the agglomerative clustering algorithm proposed by Beeferman and Berger (2000). We point out a weakness of the method caused by noisy user clicks and propose an improved clustering algorithm. Our results show that in general the agglomerative clustering algorithm can cluster similar queries effectively and that our improved algorithm can successful eliminate noisy clicks and produce cleaner query clusters.