Research On HNC Theory And Random Fuzzy And Their Application In Question Answering System

Along with the expansion of information on the Internet, it is more and more difficult to find the information that we need on the Internet. Although many engines such as Google and Baidu has provided tools to effectively search information from enormous web pages, but since modern search engines are generally based on keywords matching, there are lots of redundant or useless information in the results, thus reduce the accuracy of the result. In order to solve the problem, we propose a question and answer system which is based on HNC theory in this paper, it can not only utilize enormous information more effectively on the Internet but also make the results more abundant and more accurate.Research on question and answer system at home and abroad is not very satisfying. There are two reasons. First, the categorizing of questions is not accurate so the answer may deviate from the original question. Second, the current answer extraction approaches are usually based on statistic method, which ignores the semantic of sentences, and thus influences the accuracy of the result. To solve the problems, we focus on the research on questions categorizing and the extraction of answer candidates. First of all, we propose a Chinese question categorizing method which is based on HNC theory in order to categorize the questions more accurate. Second, we also put forward a multi-strategy answer extraction method based on HNC theory to improve the quality of answer generation.Generally speaking, this paper mainly concentrates on the following three aspects:(1) With random fuzzy theory and HNC theory, we propose a random fuzzy tree. By calculating the sentence type primitive opportunity of random fuzzy tree, we can digest the HNC fuzzy to a certain extent.(2) Put forward a Chinese question categorizing method based on HNC theory. With HNC theory and random fuzzy semantic knowledge, the various and complicated character of Chinese has been completely satisfied.(3) A multi-strategy answer extraction algorithm based on HNC is proposed in this paper. The replacement of synonyms make the representation range of answer candidates much broader, hence the calculation method of concept similarity is optimized and is combined with pattern matching. All these efforts improve the access rate and the accuracy of answer extraction.(4) In order to prove the theory proposed in the paper, we have implemented a prototype system and take Shanghaixiandai as background. Many comparison experiments is done. All the results demonstrate that the algorithms are effective.