Beyond Question Answering: Understanding the Information Need of the User

Intelligent interaction between humans and computers has been a dream of artificial intelligence since the beginning of digital era and one of the original motivations behind the creation of artificial intelligence. A key step towards the achievement of such an ambitious goal is to enable the Question Answering systems understand the information need of the user.
In this thesis, we attempt to enable the QA system's ability to understand the user's information need by three approaches. First, an clarification question generation method is proposed to help the user clarify the information need and bridge information need gap between QA system and the user. Next, a translation based model is obtained from the large archives of Community Question Answering data, to model the information need behind a question and boost the performance of question recommendation. Finally, a fine-grained classification framework is proposed to enable the systems to recommend answered questions based on information need satisfaction.