A Spatial and Sequential Combined Method for Web Service Classification

Abstract

With the growing prosperity of the Web service ecosystem, high-quality service classification has become an essential requirement. Web service description documents are semantic definitions of services, which is edited by service developers to include not only usage scenarios and functions of services but also a lot of prior knowledge and jargons. However, at present, existing deep learning models cannot fully extract the heterogeneous features of service description documents, resulting in unsatisfactory service classification results. In this paper, we propose a novel deep neural network which integrates the Graph Convolutional Network (GCN) with Bidirectional Long Short-Term Memory (Bi-LSTM) network to automatically extract the features of function description documents for Web services. Specifically, we first utilize a two-layer GCN to extract global spatial structure features of Web services, which serves as a pre-training word embedding process. Afterwards, the sequential features of Web services learned from the Bi-LSTM model are integrated for joint training of parameters. Experimental results demonstrate that our proposed method outperforms various state-of-the-art methods in classification performance.