Date of Publication : June 2026

Abstract: As the number of software repositories has greatly expanded and the amount of open-source contribution has in-creased, the challenge of being able to efficiently retrieve code has become a major issue when developing software. Developers are also frequently using search engines to help them locate reusable pieces of code, application programming interfaces (APIs), and techniques for implementation. However, most of the current solutions for this type of searching are based on traditional methods of retrieving data, on keyword input; therefore, they do not take into account either the semantic or the structural meaning of the code being searched. In addition, they are not very transparent when displaying their rank ordering of results, so that it is often difficult for developers to know why one piece of code is shown to be more important than another. This research describes a new search engine that is code aware, has explainable ranking, and combines semantic knowledge of coding with structural analysis and interpretable ranking mechanisms. The new system uses the concepts of abstract syntax tree (AST) parsing, vector embedding, and hybrid similarity measures to enhance the quality of the results produced by a search engine. The Explainable Module of the new system provides an explanation of how the results were generated, which increases the trust and usability of this system. The results of an experimental evaluation show that the new Code-Aware Search Engine with Explainable Ranking is superior to traditional keyword-based search engines and deep learning searches in terms of performance for retrieval accuracy, explainable, and relevance. In addition, the new Code-Aware Search Engine with Explainable Ranking can do this with low response times, which makes it suitable for use in real-time development environments.

Reference :