Research On Web Service Discovery Method Based On Semantic Clustering And Multi-dimensional Matching

With the rapid development of the Internet,the Service-Oriented Architecture based system design has been widely used.As the basis of SOA architecture,Web services have the loose coupling,platform independence,and realizing data exchange without the additional support of third-party software and hardware characteristics,which can provide key resources for information transmission and information sharing around the world.However,with the popularization and improvement of Web service technology,the number and types of Web services in the Internet are increasing on a large scale,generating a large number of Web services with different functions,quality and granularity.Therefore,how to quickly and accurately discover Web services that meet user's requirements from a complex service set has become a key problem to be solved in current research of Web service discovery.Based on the existing work,this thesis takes the user's functional requirements as the core and narrows the scope of service retrieval by using semantic clustering.Furthermore,this study constructs the word representation from the dimensions of word frequency,static semantic and dynamic semantic,and uses the convolution neural network to predict matching level,which aims to capture comprehensive semantic information and accurately discover the required Web service.The main contributions of our work are concluded as:(1)A Web service clustering and matching method based on functional semantic similarity is proposed.According to the functional requirements of user,the service description,service name and service I/O interface data are extracted from the semantic description file of Web service.Besides,this study takes the PV-DM model to realize the vectorization for service description.Moreover,the parameter setting and performance of three clustering algorithms based on partition,hierarchy,and density are performed and compared through the contour coefficient analysis.The most appropriate algorithm is selected to cluster the Web services.Finally,Web service categories are matched according to the overall functional semantic similarity,and the service discovery scope is thus reduced to the matched cluster.(2)A Web service discovery method based on multidimensional word representation matching is proposed.Firstly,the standard text matching dataset is preprocessed,including data filtering,keyword extraction,and multidimensional word vector representation generation from the dimension of word frequency,static semantics and dynamic semantics.Additionally,this study obtains the similarity matrix based on the multidimensional word representation to generate training input samples.Secondly,the convolution neural network is applied to construct the text matching model to learn the mapping relationship between the similarity matrix and matching tags.After that,the sample data within the matched Web service cluster is processed,and the Web service discovery is therefore transformed into a binary text classification problem(matching or mismatching).In accordance with the value of prediction probability,the candidate Web services are sorted to find the most precise Web services,which ensures the accuracy of Web Service discovery.(3)Experimental evaluation analysis and prototype development.This thesis compares the performance of this proposed method with three groups of comparison methods in terms of accuracy and error loss on the OWLS-TC4 standard service test collections.Besides,we verifies the effectiveness of combining clustering technology and multidimensional word representation matching method to perform Web service discovery from multiple perspectives.According to the process of proposed semantic clustering and multidimensional matching based Web service discovery method,the prototype is designed and developed to realize the interactive Web service discovery.