CRRA: A collaborative approach to re-ranking search results

Web search is not only an individual activity, but also a collaborative activity. By analyzing users' search activities at a community level, effectively refining the ranking of search results has received wide attention in information retrieval in recent years. Previous research has proposed approaches that predefined communities to search collaboratively. However, these approaches usually neglected correlation between users. In this study, we propose a novel collaborative approach (CRRA) for re-ranking search results based on user search activities recorded in query logs. The central idea is to establish correlations among three factors: user, query and document terms, by analyzing user logs. These correlations are then employed to determine community dynamically based on probability theory and collaborative filtering technique and calculate re-ranking score of search results. Evaluation results show that CRRA is more effective than other collaborative ranking approaches.