Entity based Q&A Retrieval

Proceedings of the 2012 Joint Conference on Empirical Methods in Natural Language Processing and Computational Natural Language Learning, pages 1266–1277, Jeju Island, Korea, 12–14 July 2012. c⃝2012 Association for Computational Linguistics Entity based Q&A retrieval Amit Singh IBM Research Bangalore, India amising3@in.ibm.com Abstract Bridging the lexical gap between the user’s question and the question-answer pairs in the Q&A archives has been a major challenge for Q&A retrieval. State-of-the-art approaches address this issue by implicitly expanding the queries with additional words using statistical translation models. While useful, the effec- tiveness of these models is highly dependant on the availability of quality corpus in the ab- sence of which they are troubled by noise is- sues. Moreover these models perform word based expansion in a context agnostic manner resulting in translation that might be mixed and fairly general. This results in degraded retrieval performance. In this work we ad- dress the above issues by extending the lex- ical word based translation model to incor- porate semantic concepts (entities). We ex- plore strategies to learn the translation proba- bilities between words and the concepts using the Q&A archives and a popular entity cata- log. Experiments conducted on a large scale real data show that the proposed techniques are promising. 1 Introduction Over the past few years community-based ques- tion answering (CQA) portals like Naver, Ya- hoo! Answers, Baidu Zhidao and WikiAnswers have attracted great attention from both academia and industry (Adamic et al., 2008; Singh and Visweswariah, 2011). These portals foster collab- orative creation of content by allowing the users to both submit questions to be answered and answer questions asked by other users. These portals aim to provide highly focused access to this information by directly returning pertinent question and answer (Q&A) pairs to the users questions, instead of a long list of ranked URLs. This is in noted contrast to the usual search paradigm, where the question is used to search the database of potential answers, in this case the question is used to search the database of pre- vious questions, which in turn are associated with answers. This involves addressing the word mis- match problem between the users question and the question-answer pairs in the archive. This is the ma- jor challenge for Q&A retrieval. Researchers have proposed the use of translation models (Berger and Lafferty, 1999; Jeon et al., 2005; Xue et al., 2008) to solve this problem. As a princi- pled approach to capturing semantic word relations, statistical translation language models are built by using the IBM model 1 (Brown et al., 1993) and have been shown to outperform traditional docu- ment language models on Q&A retrieval task. The basic idea is to estimate the likelihood of translat- ing a document1 to a query by exploiting the depen- dencies that exists between query words and doc- ument words. For example the document contain- ing the word Wheezing may well answer the ques- tion containing the term Asthma. They learn the these dependencies (encoded as translation proba- bilities) between …

Content truncated to 500 words