TranS^3: A Transformer-based Framework for Unifying Code Summarization and Code Search

TL;DR

TranS^3 is a transformer-based framework that unifies code summarization and code search, improving accuracy in both tasks through an actor-critic model and iterative feedback.

Contribution

It introduces a novel unified transformer-based framework combining code summarization and search with an actor-critic network for enhanced performance.

Findings

Significantly outperforms state-of-the-art methods in code summarization.

Achieves higher accuracy in code search tasks.

Validated by experimental and case studies with developer feedback.

Abstract

Code summarization and code search have been widely adopted in sofwaredevelopmentandmaintenance. However, fewstudieshave explored the efcacy of unifying them. In this paper, we propose TranS^3 , a transformer-based framework to integrate code summarization with code search. Specifcally, for code summarization,TranS^3 enables an actor-critic network, where in the actor network, we encode the collected code snippets via transformer- and tree-transformer-based encoder and decode the given code snippet to generate its comment. Meanwhile, we iteratively tune the actor network via the feedback from the critic network for enhancing the quality of the generated comments. Furthermore, we import the generated comments to code search for enhancing its accuracy. To evaluatetheefectivenessof TranS^3 , we conduct a set of experimental studies and case studies where the experimental results suggest that TranS^3 can signifcantly outperform multiple state-of-the-art approaches in both code summarization and code search and the study results further strengthen the efcacy of TranS^3 from the developers' points of view.