Translation Language Model Enhancement for Community Question Retrieval Using User Adoption Answer

Abstract

Community Question Answering (CQA) services on Web provide an important alternative for knowledge acquisition. As an essential component of CQA services, question retrieval can help users save much time by finding relevant questions. However, there is a “gap” between queried questions and candidate questions, which is called lexical chasm or word mismatch problem. In this paper, we improve traditional Topic inference based Translation Language Model (T\(^2\)LM) by using the topic information of queries. Moreover, we make use of user information, specifically the number of user adoption answers, for further enhancing our proposed model. In our model, the translation model and the topic model “bridge” the word gap by linking different words. Besides, user information that has no direct relation with semantics is used to help us “bypass” the gap. By combining both of them we obtain a considerable improvement for the performance of question retrieval. Experimental results on a real Chinese CQA data set show that our proposed model improves the retrieval performance over T\(^2\)LM baseline by 7.5% in terms of Mean Average Precision (MAP).