Generating Query Suggestions to Support Task-Based Search

Date: August 8, 2017
Venue: Tokyo, Japan. The 40th International ACM SIGIR Conference on Research and Development in Information Retrieval (SIGIR '17)
Corresponding article: https://arxiv.org/abs/1708.08289

Please cite the paper, and link to or credit this presentation when using it or part of it in your work.

#InformationRetrieval #IR #QuerySuggestions #TaskBasedSearch

Transcript

1. GENERATING QUERY SUGGESTIONS TO SUPPORT TASK-BASED SEARCH Darío Garigliotti and Krisztian Balog University of Stavanger Given an initial query (q0), we want to get a ranked list of query suggestions (q1, ..., qn) that cover all the possible subtasks related to the task the user is trying to achieve. Get the paper References: [1] M. Verma, E. Kanoulas, E. Yilmaz, R. Mehrotra, B. Carterette, N. Craswell, and P. Bailey. Overview of the TREC Tasks Track 2016. In Proc. of TREC '16. • This is the problem definition of the task understanding task at the TREC Tasks track - Systems are evaluated in terms of relevance and diversity of their query suggestions to the subtasks - We were the top-scoring participant in 2016 [1] choose bathroom choose bathroom cabinets lightning choose bathroom decoration style bathroom get ideas renew floor bathroom changing furniture bathroom An envisaged interface for task-based search We thank SIGIR for the Student Travel Grant … AND ANALIZE ITS COMPONENTS RQ: What are effective ways of estimating the importance of documents? • Using web query suggestions accordingly to their rank • Uniformly for web search snippets and documents RQ: What are effective ways of generating query suggestions from keyphrases? • Using the raw keyphrases, as-is, performs better than expanding them by taking the original q0 into account RQ: What are the most useful information sources? 46 Google 46 DDG W6 Bing W6 Google W6 DDG WD Bing WD Google WD DDG W H 46 Bing 0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.35 0.40 E55-IA@20 2015 4uHry sHW 2016 4uHry sHW • Followed by keyphrases extracted from (i) web search snippets (ii) web search full documents (iii) WikiHow articles • Query suggestions provided by major web search engines are the most useful information source WE PROPOSE A GENERATIVE PROBABILISTIC MODEL… Probability that a query suggestion q was generated by the initial q0: P(q|q0) = X s X d ⇣ X k P(q|q0, s, k)P(k|s, d) ⌘ P(d|q0, s) P(s|q0) Query Suggestion Generation Document Importance Keyphrase Relevance Source Importance QS WS WD WH Query suggestions Keyphrases q0 q0 Architecture of our system Test dataset excerpt for initial query: "low wedding budget" Cheap wedding cake Make your own invitations Buy a used wedding gown } } } Relevance Subtask ID ... 7 1 low budget wedding dresses 7 0 low wedding budget cars 7 1 find a gown 8 0 wedding flowers 8 1 cup cake wedding 8 1 wedding cakes 9 2 wedding invitation 9 1 find wedding invitation templates 9 0 designer dresses wedding ...