An intent-enhanced feedback extension model for code search

Abstract

Context:

Queries and descriptions used for code search not only differ in semantics and syntax, but also in structural features. Therefore, solving the differences between them is of great significance to the study of code search.

Objective:

This study focuses on the improvement of code search accuracy by exploring the expansion of query statements during the search process.

Methods:

To address the disparities between description and query, the paper introduces the Intentional Enhancement and Feedback (QEIEF) query expansion model. QEIEF leverages the written description provided by developers as the source for query expansion. Furthermore, QEIEF incorporates theQEIEF method to enhance the semantic representation of the query. This involves utilizing the query output as the target for intent enhancement and integrating it back into the query.

Results:

To assess the effectiveness of the proposedQEIEF in code search tasks, we conducted experiments using two base models (DeepCS and UNIF) along withQEIEF, as well as baseline models (WordNet and BM25). The experimental results indicate that QEIEF outperforms the baseline models in terms of query expansion accuracy and code search results.

Conclusion:

QEIEF not only enhances the accuracy of query expansion but also substantially improves code search performance. The source code and data associated with our study can be accessed publicly at: The address of our new code and data is https://github.com/xiangzheng666/IST-IEFE.