Web API Discovery Using Semantic Similarity and Hungarian Algorithm

Abstract

Representational state transfer (REST) is the current design standard for Web application programmable interfaces (APIs). Unfortunately, existing Web API search engines allow for only keyword or tag-based searches. Furthermore, search engines do not take into account the semantics and the interface compatibility. This paper reports on a novel approach to RESTful service discovery, referred to as Interface-Compatibility-based Semantic Service Search (ICS^3). ICS^3 expands terms in service documents using DBpedia and WordNet, and then filters out services that are inapplicable to the user query, in two steps: (1) calculating semantic similarity between candidate services and the user query and (2) calculating the degree of interface compatibility between candidate services and the user query using the Hungarian algorithm. Experiment results demonstrate that ICS^3 outperforms other Web APIs search methods in terms of accuracy.