Ranking Entities Using Web Search Query Logs

Searching for entities is an emerging task in Information Retrieval for which the goal is finding well defined entities instead of documents matching the query terms. In this paper we propose a novel approach to Entity Retrieval by using Web search engine query logs. We use Markov random walks on (1) Click Graphs – built from clickthrough data – and on (2) Session Graphs – built from user session information. We thus provide semantic bridges between different query terms, and therefore indicate meaningful connections between Entity Retrieval queries and related entities.