Automated Web Service Discovery and computing from public repositories through probabilistic matchmaking

Abstract

This Paper is an attempt to develop an efficient and comprehensive approach to web service discovery and retrieving APIs from vast repositories based on user queries and requirements and ranking the results based on relevance. Our strategy to do so incorporates numerous techniques like semantic search, graph-based ranking, and relevance scoring. Our module applies semantic expansion on user queries the moment they are received, through WordNet to improve query representation. The next step is to vectorize the expanded query through TF-IDF, which facilitates semantic similarity computation with the web services available. The semantic similarity scores are then studied with the help of a graph where the edges are semantic similarity scores and the nodes represent web services. Importance scores are then given to each web service on this graph with the help of PageRank, helping us understand the relevance of the web services. Not just this, the Okapi BM25 algorithm is also applied to compute the relevance score. The final ranking of the web service is given on the basis of integrated scores of Okapi BM25 and PageRank. This ranked list is finally presented to the user. With the help of our module and the approach it follows, users can navigate through vast repositories full of APIs to find the most relevant API for their use. Through the approach followed by us, web service discovery and ranking becomes easier even for people without a lot of experience and hence it offers a strong and effective solution to web service discovery and can be applied in multiple domains.