| The question answering system is able to quickly return an accurate and concise answer to the questions raised by the users in natural language,rather than a series of web pages related to the questions.Compared to the general search engine,the question answering system reflects its simple and clear characteristics.Semantic similarity computation is a key technique in the question answering system.The accuracy of the computation method has a direct effect on the performance of the question answering system,and it is also a hot spot for research.The semantic similarity computation in the question and answer system is divided into two levels,the semantic similarity computation at the concept level and the semantic similarity computation at the sentence level.Firstly,in order to solve the problem that the current concept semantic similarity computation methods do not fully consider the semantic information carried by the concept nodes in the ontology,we present an improved concept semantic similarity computation method based on HowNet ontology(BESM).BESM is divided into two steps,the first step is to compute the similarity between sememes and the second step is to compute the similarity between concepts.Where,the semantic similarity between sememes is weighted by the basic semantic similarity and the adjustment factors.The experimental results show that BECM perfects the factors needed to be considered in the concept semantic similarity computation;the Pearson correlation between the similarities computed by BECM between expert experiences ups to 95.1%and BECM has a high accuracy.Secondly,on the basis of the concept semantic similarity calculation method,we present a sentence semantic similarity calculation method.This method builds a set of candidate questions through the FAQ rewind index structure to reduce the size and consumption of the calculation.Then through the semantic similarity between the concepts to obtain the semantic similarity between the keywords,and final get the semantic similarity between the sentences.Finally,in order to verify the performance of the semantic similarity calculation method,we designed and implemented a FAQ question answering system based on Premier League knowledge.This system is used to answer the field related questions raised by the users.The experimental results show that the system has a high response speed,accuracy rate and recall rate,and then verify the performance of the similarity method. |
