Tempo-Lexical Context Driven Word Embedding for Cross-Session Search Task Extraction

Proceedings of NAACL-HLT 2018, pages 283–292 New Orleans, Louisiana, June 1 - 6, 2018. c⃝2018 Association for Computational Linguistics Tempo-Lexical Context driven Word Embedding for Cross-Session Search Task Extraction Procheta Sen ADAPT Centre School of Computing Dublin City University Dublin 9, Ireland procheta.sen2@mail.dcu.ie Debasis Ganguly IBM Research Labs Dublin, Ireland debasis.ganguly1@ie.ibm.com Gareth J.F. Jones School of Computing ADAPT Centre Dublin City University Dublin 9, Ireland Gareth.Jones@dcu.ie Abstract Search task extraction in information retrieval is the process of identifying search intents over a set of queries relating to the same topical in- formation need. Search tasks may potentially span across multiple search sessions. Most ex- isting research on search task extraction has focused on identifying tasks within a single session, where the notion of a session is de- fined by a fixed length time window. By con- trast, in this work we seek to identify tasks that span across multiple sessions. To iden- tify tasks, we conduct a global analysis of a query log in its entirety without restricting analysis to individual temporal windows. To capture inherent task semantics, we represent queries as vectors in an abstract space. We learn the embedding of query words in this space by leveraging the temporal and lexical contexts of queries. To evaluate the effec- tiveness of the proposed query embedding, we conduct experiments of clustering queries into tasks with a particular interest of measuring the cross-session search task recall. Results of our experiments demonstrate that task extrac- tion effectiveness, including cross-session re- call, is improved significantly with the help of our proposed method of embedding the query terms by leveraging the temporal and temp- lexical contexts of queries. 1 Introduction A complex search task is defined as a “a multi- aspect or a multi-step information need consisting of a set of related subtasks, each of which might recursively be complex” (Hassan Awadallah et al., 2014). For example, a task of making arrange- ments for travel to a conference qualifies as a com- plex search task because there are several choices that a user needs to make in order to plan his entire trip, e.g. selecting flight, hotel, making arrange- ments for local transport, finding the conference venue, finding good places to eat around, finding local sight-seeing options after the conference etc. All these sub-tasks are likely to take place within their own search sessions, where a session is de- fined as a set comprised of queries executed dur- ing a time period of a specific length, usually about half-an-hour (Lucchese et al., 2013). In this paper, we address the problem of au- tomatically predicting whether search sessions, focused on specific activities, are a part of a broader complex search task, which we refer to as the cross-session search task extraction prob- lem. Cross-session search task extraction can po- tentially find applications in designing more pro- active search engines, which may suggest rele- vant information about specific subsequent sub- tasks along a timeline, e.g. suggesting places to eat around a conference venue without the user …

Content truncated to 500 words