API Prober – A Tool for Analyzing Web API Features and Clustering Web APIs

Abstract

Nowadays, Web services attract more and more attentions. Many companies expose their data or services by publishing Web APIs (Application Programming Interface) to let users create innovative services or applications. To ease the use of various and complex APIs, multiple API directory services or API search engines, such as Mashape, API Harmony, and ProgrammableWeb, are emerging in recent years. However, most API systems are only able to help developers to understand Web APIs. Furthermore, these systems do neither provide usage examples for users, nor help users understand the “closeness” between APIs. Therefore, we propose a system, referred to as API Prober, to address the above issues by constructing an API “dictionary”. There are multiple main features of API Prober. First, API Prober transforms OAS (OpenAPI Specification 2.0) into the graph structure in Neo4J database and annotates the semantic concepts on each graph node by using LDA (Latent Dirichlet Allocation) and WordNet. Second, by parsing source codes in the GitHub, API Prober is able to retrieve code examples that utilize APIs. Third, API Prober performs API classification through cluster analysis for OAS documents. Finally, the experimental results show that API Prober can appropriately produce service clusters.