Automatically Generating Questions from Queries for Community-based Question Answering

Proceedings of the 5th International Joint Conference on Natural Language Processing, pages 929–937, Chiang Mai, Thailand, November 8 – 13, 2011. c⃝2011 AFNLP Automatically Generating Questions from Queries for Community-based Question Answering Shiqi Zhao1,2, Haifeng Wang1, Chao Li2, Ting Liu2, Yi Guan2 1Baidu, Beijing, China {zhaoshiqi,wanghaifeng}@baidu.com 2Harbin Institute of Technology, Harbin, China beyondlee2008@yahoo.cn, tliu@ir.hit.edu.cn, guanyi@hit.edu.cn Abstract This paper proposes a method that auto- matically generates questions from queries for community-based question answering (cQA) services. Our query-to-question generation model is built upon templates induced from search engine query logs. In detail, we first extract pairs of queries and user-clicked questions from query logs, with which we induce question generation templates. Then, when a new query is submitted, we select proper templates for the query and generate questions through template instantiation. We evaluated the method with a set of short queries ran- domly selected from query logs, and the generated questions were judged by hu- man annotators. Experimental results show that, the precision of 1-best and 5- best generated questions is 67% and 61%, respectively, which outperforms a baseline method that directly retrieves questions for queries in a cQA site search engine. In addition, the results also suggest that the proposed method can improve the search of cQA archives. 1 Introduction In recent years, community-based question an- swering (cQA) services become popular, such as Yahoo! Answers (answers.yahoo.com) in English and Zhidao (zhidao.baidu.com) in Chinese. In cQA, people can have their questions answered by other people rather than by automatic QA sys- tems, which usually better guarantee the answer quality. Till Oct. 2010, Zhidao has accumulated over 100 million answered questions, which form an extremely large and valuable knowledge base. Lin (2008) first proposed the idea of automati- cally generating questions from queries. The un- derlying assumption is that when a user issues a query to a search engine, he could have a question in mind but it is more convenient and efficient for him to realize the question as a query. This tech- nique could have a great impact on cQA services. First, it can improve the search of cQA archives. As we know, most of cQA resources can be searched with general search engines like Google and Baidu (www.baidu.com). Many of them also have their own site search engines. However, since user queries are mostly short and incomplete, it is quite often that many less relevant questions are retrieved when searching cQA archives. By gener- ating questions from short queries, we can expand the queries and estimate questions that are more likely to be interested in, which could help to re- trieve more related questions from cQA archives. Second, this technique can be useful in enlarg- ing cQA resources. For example, in some search engines like Baidu, if a query is found to be fre- quently searched, the query will be automatically submitted to a cQA site and expected to be an- swered by some users. However, a problem is that the frequent queries are usually short and incom- prehensible, which …

Content truncated to 500 words