Automatic Identification Of Problematic Situations For Interactive Question Answering System

With the continuous development of internet, it is not satisfied to get informationby using the search engine only. Many researchers focus on how to provide theimportant information for us quickly and easily. Question answering system (QA system)has the two aspect features: one is to satisfy the users’ need for information, and theother one is to provide the humanity answers to questions. AlthoughQA system cansolve the problem very well, it has shortcomingsthatconversional QA system has nomechanism to identify the problematic situations automatically.In order to compensate for these shortcomings of conventional QA system, it isessential to research on the identification of problematic situations forinteractive QAsystem.In the paper, an approach of the identification of problematic situations ofinteractive QA system is proposed. It is necessary to analyze the identification ofproblematic situations which based on the users’ emotion, intent and mixturefeatures.The efficiency is raised because the method to identify the problematicsituations is changed from manual review to identify automatically.In order to identify the problematic situations better than before, the method toexpanse the knowledge base is designed. The magnitude the knowledge base for QAsystem has expanded from2069to39161. They satisfy the users’ need approximatelywith the different kinds of corpus in different aspects.On this basis, the architecture of QA system has been improved than before. AndBIT robot QA system has three different platforms which contain the wechat version,the QQ version and the web version. They are the reasons why the QA system attractedso many users to use it.After identifying these problematic situations, manual review is the way to improvethe error answers. Finally, the correct answers should be updated to the knowledge baseof QA system.During the experiment, the corpus comes from the wechat platform of QAsystem.it is essential to callout and analyze the corpus and then to propose theapproaches to identify problematic situations. The precision of the identification ofproblematic situationsis up to76.77%. The experimental results and theoperating of QAsystemverified the effectiveness of the proposed approach.