Web Service Clustering Technique based on Contextual Word Embedding for Service Representation

Due to extensive use of Internet and IoT, the demand Web services and APIs are increasing day by day and there is a proliferation of services on internet in terms of quality and quantity both. It raises the need of service management. Web service clustering plays a vital role in service management as it reduces the search space and time. Word2vec word embedding is highly demanded in these days as it can capture the semantic similarity but it does not bother the context and it effects the clustering performance. In this paper, Sentence-BERT (Sentence Bidirectional Encoder Representations from Transformers) embedding is used in the vector space representation of services so that with the semantic meaning, context of the features can be also analyzed and services can be efficiently mapped in vector space. To analyze the performance of embedding, K-Means clustering is applied and results are compared with the different state-of-art techniques based on standard evaluation measures. The experimental results shows that accuracy of the proposed model is increased by approximately 49% in comparison of a model in which word2vec model is utilized.