Web Service Clustering Method Based on Word Vector and Biterm Topic Model

摘要

The increasing number of Web services brings challenges for accurate and efficient service discovery. Service clustering technology can narrow the retrieval range of service discovery and improve the efficiency of service discovery. However, due to the short text features with few words and difficult feature extraction in service description text, the traditional text clustering model can not get effective clusters. In this paper, a web service clustering method based on word vector and BTM (Biterm Topic Model) is proposed. Firstly, the sparse feature problem of short text is alleviated by expanding the word vector. Secondly, the extended service description text is modeled by the BTM based on Gibbs sampling. Finally, the k-means algorithm is used to cluster web services. Compared with LDA (Latent Dirichlet Allocation) and BTM, this method improves the F-Measure of clustering index. Experimental results show that the proposed method can effectively alleviate the problem of sparse service description text features and improve the clustering effect of services.