A Method to Enhance Web Service Clustering by Integrating Label-Enhanced Functional Semantics and Service Collaboration

Abstract

In Web service clustering, the service function vector (SFV) directly determines the quality of service (QoS) clustering. To improve service clustering performance, a method is proposed in this paper by integrating label-enhanced functional semantics and service collaboration. It improves the SFV from three aspects: generation model, corpus, and structural auxiliary information. At the generation model level, a Sentence-BERT is constructed based on singular value decomposition (SVD), to alleviate the anisotropy problem of BERT in vectorizing service descriptions. For corpus, the semantic features of SFV are supplemented by extracting specific named entities from service descriptions. Meanwhile, the service collaboration graph is established according to the collaboration relationship among Web services, which is conducive to the variational graph auto-encoders (VGAE) to realize service collaboration feature aggregation and further improve the SFV. Experiments show that the improved model, corpus and structural auxiliary information effectively enhance the SFV clustering. The proposed Web service clustering method is superior to the state-of-the-art methods.

Keywords

• Service collaboration
• service description
• service function vector (SFV)
• variational graph auto-encoders (VGAE)
• web service clustering