Domain Knowledge Domain Question Answering System Answers Extracted

Question Answering System can offer people the natural language inquiry and return the answer directly other than the massive WebPages. Compared with traditional search engines, Question Answering System can better express the users' need, adapt to the users' habits and the answer information is also much quicker, more accurate and efficient. It can overcome the disadvantages of traditional search and is the hot topic of current research. However, due to the inherent complexity of natural language and technical and other reasons, the accuracy of the current territory general Question Answering System is far to meet the practical requirements. Restricted domain Question Answering System serves designated areas and relies on specific domain knowledge. The restriction of service domain and the structure of knowledge base are relatively in good order, which reduces the difficulty of dealing with natural language to some extent. Thus, the research difficulty of restricted domain Question Answering System is reduced and it is possible to put it into practical use soon.This paper probes into the key techniques of restricted domain Question Answering System from some aspects, namely how to filter out irrelevant information from the mass of network resources in the domain with existing domain knowledge and how to extract the accurate answers from the obtained domain information by analyzing the demand of users. The progress has been made in the following aspects.(1) Proposing the approached of domain text classification model. This method adopts support vector machine learning algorithm, combines sample statistics with domain terms to form domain classification feature space, and calculates relevance of domain concepts using the relations of domain interior knowledge. Thus, certain weigh is given to the classified characteristic and domain text classification model is formed. The use of domain text classification model to filter non-domain text provides the Question Answering System with a great deal of domain information as well as raises the accuracy of text retrieval.(2) Proposing domain passage segmentation method and domain passage retrieval algorithm. With the help of the thought of passage retrieval, this method combines domain question feature and domain answers to the questions and proposes the passage segmentation fit for domain questions. And this method improves passage retrieval algorithm based on the density with domain knowledge and extract the most related passages to be candidate passage, further reduces the answer extraction scope.(3) Proposing answer-extraction strategy and the algorithm based on the domain characteristic. To simple fact question and definition question in the restricted domain Question Answering System, we adopt algorithm of the key word distribution density and similarity computation of question and answer. To enumerative question in the restricted domain Question Answering System, we use naming entity recognition technology based on the algorithm Hidden Markov Models and the Condition Random Field and raise recalling rate and rate of accuracy of answer extraction.(4) Treating YunNan tourism as restricted domain to do YunNan tourism text retrieval and classification experiment, passage retrieval and answer extraction experiment, design and realize Question Answering System of YunNan tourism.