Research On Sentence Semantic Similarity Based On WordNet In Automatic Question Answering System

In order to improve the retrieval quality, a new information retrieval method--automatic question answering system is proposed which can answer questions raised by users fast and succinctly, it is currently a hot research direction, and has much room for development. One of its key technologies in QA system is calculation of sentence similarity, the quality of sentence similarity algorithm affects its performance directly. At present, there are many studies of sentence similarity among which the sentence semantic similarity algorithm is a research hotspot.This paper studies the sentence semantic similarity algorithm in QA system based on WordNet. WordNet is a large English vocabulary database which is used widely in the field of natural language processing, information retrieval and other applications. Vocabulary is the basic element of sentence and sentence semantic information contained in the word semantics. In WordNet the word semantic is represented by the concept, so the core of the sentence semantic similarity algorithm can be attributed to computing concept semantic similarity. Through a large number of literature and algorithm research, the related factors which affected concept semantic similarity algorithm in WordNet are analyzed and a new CP weighted method is proposed by using the semantic hierarchy among concepts which consists of is-a relations in WordeNet, The method takes the probability of one concept appearing when another concept appearing(conditional probability) as a parameter to measure the weight of hyponymy between concepts. The weights of hypernym and hypogyny are distinguished which make the weight distribution more reasonable. An improved model of concept semantic similarity algorithm is given based on the method. The model is not only considers the concept of the density and depth of the classification tree, but also takes the inter path between the two concepts into consideration, so that the accuracy of semantic similarity between concepts has been improved. The model is applied into the sentence semantic similarity algorithm to improve the algorithm performance. In this paper, the main work includes:Firstly, this paper expounds the research background and significance, then introduce the research status of QA systems, sentence similarity algorithm and Concept semantic similarity.Secondly, a brief introduction to the knowledge of QA systems and common sentence similarity algorithm is given.Then, according to the semantic hierarchy in WordNet, a new and improved method of weighting algorithm model is put forward, the improved concept similarity algorithm model is applied into sentence semantic similarity algorithm and is verify by experiments.Finally, summarize the work of this article and point out the problem needing to be solved further as well as the research direction in the future.