Interactive Question Answering System Based On Hybrid-structure Data Sources

With the development of the Internet, information resources grow fast, forming a huge, rich content information database. The traditional information retrieval system, users input keywords to search, the system returns some relevant web pages of information, requires the user to find relevant information from these answers. General Q&A information retrieval although the user can direct get the answer, but users can not conduct in-depth interaction with the system, the problem can not be continuous. Based on the real question answering of the interactive Q&A system, it allows users to ask questions in natural language, the system gives a direct answer to the question, the user can be given in-depth questions according to the answer, or raise new issues.This issue builds an interactive question answering system based on hybird-structure data sources, mainly to solve three problems:A.Information retrieval based on hybrid-structure data sources:This issue has build a Q&A knowledge base, a real-time database, and a common-sense knowledge database. Q&A knowledge base is updated regularly each day to get new questions and answers, real-time database stores the current real-time data information and common-sense knowledge base stores common sense knowledge. Interactive Q&A system, retrieval answers from these three databases.B. Multi-level user information analysis:For the interactive Q&A systems, understand the question is the key point to answer the question. This paper use classification method to understand the complex questions. It classified complex questions by question surface information, user real intentions and user expectations three levels. By use the classification of questions, system can accurately identify the needs of users according to the questions type information at all levels, and based on user needs to find the answer to meet the requirements to improve the system.C: Context relevant question series handling:User can use natural language to ask question in Interactive Q&A system, in the course of the dialogue, user may omit topic or focus which can be find on the upper contexts. Based on the questions topic, focus, interrogative words, question omitted types are divided into nine categories, and according to the question ellipsis type and features, use decision tree classifier to recover the question, realizing the interactive chat between the users and the system.