カンファレンス (国内) Finding and Suggesting Alternative Verbal Queries from Community Q&A Corpus
第9回データ工学と情報マネジメントに関するフォーラム (DEIM 2017)
Web searchers often use aWeb search engine to find a way or mean to achieve his goal. For example, for a user who issues a query “sleeping pills,” intending to solve her sleep problem, there exists another solution to achieve her goal such as “have a cup of hot milk” or “take a stroll before bedtime.” The problem is she may not be aware that these solutions exist, so she will probably choose to take a sleeping pill without considering these solutions. In this work we define and tackle the alternative action mining problem, where a system is required to find alternative actions for a given query. We define alternative actions as actions who share the same goal and define the alternative action mining problem as similar in the search result diversification. To tackle the problem, we propose to leverage a community Q&A (cQA) corpus for mining alternative actions. The cQA corpus can be seen as an archival data of dialogues between questioners, who want to know the solutions for their goal, and answerers, who suggest different solutions for achieving it. We propose a method to compute how two actions can be alternative to each other by using a question-answer structure in cQA. Our method builds a question-action bipartite graph and recursively compute how likely two actions can be alternative to each other. We conducted the experiments to investigate the effectiveness of our method with two newly-built test collections, each of which contains 50 queries. The experimental results showed that our proposed method outperformed the query suggestions provided by commercial search engines in terms of D#-nDCG.