Proactive identification of query failure

When a user fails to find any useful information to support the task at hand after issuing a query, the user experiences a query failure. Since users possess limited cognitive resources, query failures often lead to user frustration as no clear benefit is obtained from the associated search interactions. Therefore, to improve users' search experiences, we conducted a controlled‐lab study with 40 participants, seeking to explore the extent to which query failures can be proactively identified before users start examining the retrieved results. Specifically, based on the data collected from 693 query segments generated in 80 search sessions, we used past search behaviors and current query attributes as features to build classifiers and examined the performance in capturing query failures. We report that (1) analytics algorithms utilizing past search behavioral data have significantly better performances than the baseline model in tasks of different types, and (2) The knowledge of users' search intentions can help improve the performance of the prediction model. Results pave way for developing proactive system supports for task‐based search interactions.