An Open-source Repository Retrieval Service Using Functional Semantics for Software Developers

Abstract

Software developers are encouraged to reuse mature third-party code repositories to accelerate their developing. The first step is to find some code repositories which satisfy developers’ demand for specific functionality. Repository hosting platforms, such as GitHub, NPM, etc. are good search databases. However, these platforms provide little retrieval function based on repositories’ functional semantics, which is not conducive for developers to get desired code repository by claing what functionality they require.To solve this problem, we propose and implement an open- source repository retrieval service. First, to extract structured functional semantics of open-source repositories, we design a NER (Named Entity Recognition) model to extract features of six different dimensions related to software development. Next, we construct a HIN (Heterogeneous Information Network) to represent nonstructured semantic information of repositories. Finally, by combining two modules preceding, we implement an open-source repository retrieval service, in which a software developer only needs to tell the service what functionality he (or she) wants and will get a ranked list of repositories as return. We further conduct experiments and the results demonstrate the effectiveness of the service.