Research On Text Semantic Similarity Algorithm In Intelligent Question Answering System

With the rapid development of network and information technology,the amount of data is exploding.People want to find answers to their questions in a large amount of data quickly and accurately.Under this premise,the intelligent question answering system has developed rapidly in order to meet the requirements of people to obtain useful information,which has great practical value.Semantic similarity calculation is a very basic and critical issue in the intelligent question answering system.It has been a hotspot and a difficult point in people's research in the past few years.However,due to the features of Chinese colloquialism and flexible word combination,the results of semantic similarity calculations have been far too poor,what greatly limit the development of the question and answer system.In view of the above problems,this paper has carried out research on relevant content.The main research contents and innovations are summarized as follows:(1)Aiming at the problem of low accuracy of semantic similarity calculation in intelligent question answering system,this paper proposes a semantic similarity algorithm based on dependency syntax analysis and vector space model.The algorithm first establishes a dependency syntax tree centered on the predicate verb and obtains the structural information of the sentence.Then,it combines the vector space model to map the word sequences obtained after sentence segmentation into the vector space of fixed dimension.The semantic information of the words is expressed,and finally the semantic similarity between the two sentences is obtained by using the text similarity calculation formula.The experimental results show that the proposed method can improve the semantic similarity calculation of sentences.(2)In order to solve the dimensional disaster of the vector space model and the semantic division of words,this paper introduces the Word2vec model and combines the characteristics of Chinese text to propose a semantic similarity algorithm based on Word2vec.Firstly,the algorithm uses Word2vec training to obtain a fixed-dimensional char embedding and word embedding.And then it combines the char embedding with the word embedding in a shallow fusion with the actual scene sentence text.Finally,the experimental results show that the algorithm can improve the accuracy of semantic similarity calculation effectively,in turn,the performance of the question and answer system has been improved.(3)Aiming at the problem that Word2vec can't get the semantic information of sentence hierarchy,a semantic similarity algorithm combining deep learning model framework is proposed.Firstly,the algorithm uses the Siamese Network framework to establish two identical and shared weight network structures to encode the sentences in a bidirectional serialization,fusing the word order information contained in the sentences.Then the latest self-attention mechanism proposed by Google is introduced to obtain the semantic information association between sentence words.Finally,using the operation of the vector proposed by the CompareAggregate framework,the semantic difference of each dimension of the word embedding is obtained.After experimental testing,the method obtained an F value of up to 82.35%in sample one.(4)Based on the research of text semantic similarity algorithm of intelligent question answering system,a set of intelligent question answering system for home appliance customer service is designed.Including client design,communication protocol design,server design,database and so on.