Abstract
????tper, we introduce a query suggestion approach that exploits users' search context and search logs. For a given search log, we integrate three pieces of wisdom embedded in the search context: consecutive queries, reformulation patterns between consecutive queries, and clicked URLs. When providing suggestions online, we extract concepts that represent the user's intent and associate these concepts with wisdom attained from past users who had similar search intents. Finally, customized suggestions are provided according to the current user's search pattern. The experimental results demonstrate that the proposed approach outperforms existing query suggestion methods and effectively provides users with more accurate suggestions to help them get required information faster.