02-09-2017, 04:35 PM
Users are performing increasingly complex goal-oriented tasks on the web, such as making travel arrangements, managing finances, or planning purchases. To this end, they often divide the tasks into a few codependent steps and multiple queries around these steps over and over for long periods of time. To better support users in their long-term information missions on the web, search engines track their queries and clicks while searching online. In this work, we study the problem of organizing the historical queries of a user in groups in a dynamic and automated way. Automatic query group identification is useful for a number of different search engine components and applications, such as query suggestions, results sorting, query modifications, sessions, and collaborative search. In our approach, we move beyond approaches that rely on textual similarity or time thresholds, and propose a more robust approach that leverages search query logs. We experimentally study the performance of different techniques, and show their potential, especially when combined.