Web Service Recommendation Based on Word Embedding and Topic Model
Ting Chen, Jianxun Liu, Buqing Cao, Zhenlian Peng, Yiping Wen, Run Li · 2018
Web service recommendation faces the following problems: large-scale number of Web services, the diversity of service categories, and the sparseness of service texts on the Internet. To this end, a Web service recommendation method that combines word embedding with topic models is proposed in this paper. Firstly, the English Wikipedia is used to construct a high-quality Word embedding model to obtain semantic similar words, and based on the DMM model, the GPU promotion strategy is integrated into the topic derivation process to obtain more effective service implied topics. Secondly, the obtained topic distribution vector is used to calculate the similarity. Multi-dimensional features such as similarity of Web APIs, similarity of Mashups, co-occurrence and popularity of Web APIs, are modeled by deep factorization machine to rank and predict Top-N Web APIs recommended set. Finally, experiments conducted on a real dataset, by comparing the proposed method with multiple existing Web service recommendation methods, show that the proposed approach achieves better performance.