Mashup Service Clustering Via Incorporating Functional Semantics And Synergistic Relationships

Web service is a kind of application program module that can be invoked through the network,which is widely used in software development and deployment of service-oriented architecture(SOA).In order to facilitate invocation and integration,web services are usually designed with small function granularity.Mashup service is an emerging service organization form.It is usually composed of multiple related web services,which can provide rich business functions and access to multiple types of data resources.With the wide use of SOA architecture,the number of web services and mashup services is increasing rapidly.Similar to web services,the emerging mashup services have also been widely used by developers.As a functional complex,mashup services are more difficult to discover than ordinary Web services.Service clustering is an effective method to improve service search efficiency.Traditional service clustering method relies on service function description to extract service function features,and then realizes clustering with the help of function feature similarity.However,due to short service description text and model generation performance,the quality of service function vector generated by this method is not high.Therefore,reasonable optimization of the corpus and model for generating service function vector can improve the generation quality of service function vector.In addition,incorporating the cooperative relationship between mashup services and their component Web services into service clustering can further improve the quality of clustering.Therefore,we propose a Mashup service clustering method that integrates functional semantics and synergic relationships.The main work and contributions are as follows:(1)A Sentence-BERT model with vector space correction is established,and a service function vector generation method with feature word semantic enhancement is proposed.Entity nouns are extracted from the service descriptions.These entity nouns together with the service labels constitute the functional feature words of mashup services,which are used to generate feature word vectors.A Sentence-BERT model with vector space correction is established to generate high-quality service description text vectors for Mashup service descriptions.The concatenated feature word vector and the service description text vector are used as the service function vector.This method improves the generation quality of service function vectors from the perspectives of generated corpus and model.(2)A service synergistic network is established,and a service representation vector generation method is proposed by fusing functional semantics and synergistic relationships.The service synergistic network is built according to the collaboration relationship between Mashup service and its constituent services.The MS＿GCMC(Mashup Services Graph Convolutional Matrix Completion)model is constructed,and the link prediction task is used to learn the synergistic characteristics between service nodes on the service collaboration network and generate the service representation vector.This method realizes the aggregation of mashup service functional semantics and synergistic relationships,which is helpful to improve the quality of service clustering.(3)A service clustering based on mashup service representation vector and KMeans++ is designed.The experimental results on the real mashup service and web service data sets of Programmable web website showthat the proposed method in this paper improves 58.6%,41.4%,34.7%,30.9%,18.0%,and 12.9% in ARI,AMI,NMI,Purity,SC,and DBI compared with the benchmark methods,which verifies the advancedness of the proposed clustering method.