Diversity-driven automated web API recommendation based on implicit requirements

Abstract

With the rapid growth of the web API sharing community, Mashup development has become a popular way for software developers to quickly create Mashup applications. Mashup development allows developers to quickly implement the functionality they want by combining a chosen set of web APIs, which greatly improves development efficiency. However, the existence of numerous web API candidates make it difficult to select the appropriate API quickly. Most existing automated web API recommendation approaches focus on the developer’s description of requirements but typically overestimate the developer’s ability to fully describe their application’s requirements, thus ignoring their implicit and diversified functional requirements. In this paper, we propose a novel automated approach called DI-RAR to retrieve and recommend web API groups for Mashup creation. Specifically, we formulate an automated web API recommendation task as a nondeterministic polynomial problem. First, a self-attention model assigns weights to query to distinguish the core and non-core requirements. then Dynamic planning retrieval generates steiner trees to retrieve API groups and uncovers strongly related implicit requirements to enrich the mashup’s functions. Finally we apply simhash technique to filter similar results and finally provide diverse API groups. experiments based on a real-world dataset are performed to demonstrate the feasibility and efficiency of DI-RAR.