A Service Classification Method Based on Symbiotic Attention Representation Learning

Abstract

The present invention relates to a service classification method based on symbiotic attention representation learning. Specifically, it includes the following steps: selecting a web service, which includes a service description and a service name; constructing a description service matrix by using the service description; extracting information words from the service description to construct a service information word feature matrix; using the service name to construct the service name feature matrix; fuse information word feature matrix and name feature matrix to obtain enhanced data feature matrix; use enhanced data feature matrix and service description feature matrix to establish service feature correlation matrix; obtain required vector parameters by calculating correlation matrix, and finally according to vector parameters and features The matrix derives the service class of the web service. The method of the invention extracts the service description as an additional service feature, and combines the extracted information word and the service name to form a service enhancement feature for service classification, thereby improving the classification accuracy.

Description

technical field [0001] The invention relates to the field of service classification methods, in particular to a service classification method based on symbiotic attention representation learning. Background technique [0002] With the widespread adoption of Service Oriented Architecture (SOA), web services such as Mashups are becoming more and more popular in the web platform and mobile application market. So far, there have been many works on service classification, most of them mainly focus on using keywords to match keywords in other service descriptions to measure the semantic distance between different services, the classification results will have similar Functional services are classified into the same category; these keyword-based approaches mainly rely on the quality of keywords in the service description, which are manually specified by the service provider, but because service providers have different services for similar functions It is difficult for a provider ...

AI Technical Summary

Problems solved by technology

However, these deep learning models still have some problems in service classification: first, these models usually only extract features from service descriptions to classify Web services; second, these models cannot solve the problems of data sparsity and context irrelevance

Although ServeNet-BERT has advantages in service classification, it still has limitations: ServeNet-BERT just combines features with service names and service descriptions, without fully exploiting the potential semantic correlation between service names and service descriptions

Images