Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Retrieval-Augmented Generation for Code Summarization via Hybrid GNN

Shangqing Liu, Yu Chen, Xiaofei Xie, Jingkai Siow, Yang Liu

TL;DR

This work tackles automatic code summarization by bridging retrieval-based and generation-based approaches through a retrieval-augmented Hybrid Graph Neural Network (HGNN). HGNN constructs a retrieval-augmented static Code Property Graph and a structure-aware attention-based dynamic graph, enabling Hybrid Message Passing that captures both local and global code semantics. A retrieval-informed decoder and an attention-based LSTM generate succinct natural language summaries, with significant improvements on the CCSD and PCSD benchmarks, including BLEU-4, ROUGE-L, and METEOR improvements over prior methods. The approach introduces a challenging 95k+ function dataset (CCSD) and demonstrates strong generalization, human-judged quality, and qualitative case studies, highlighting its practical impact for code understanding and maintenance.

Abstract

Source code summarization aims to generate natural language summaries from structured code snippets for better understanding code functionalities. However, automatic code summarization is challenging due to the complexity of the source code and the language gap between the source code and natural language summaries. Most previous approaches either rely on retrieval-based (which can take advantage of similar examples seen from the retrieval database, but have low generalization performance) or generation-based methods (which have better generalization performance, but cannot take advantage of similar examples). This paper proposes a novel retrieval-augmented mechanism to combine the benefits of both worlds. Furthermore, to mitigate the limitation of Graph Neural Networks (GNNs) on capturing global graph structure information of source code, we propose a novel attention-based dynamic graph to complement the static graph representation of the source code, and design a hybrid message passing GNN for capturing both the local and global structural information. To evaluate the proposed approach, we release a new challenging benchmark, crawled from diversified large-scale open-source C projects (total 95k+ unique functions in the dataset). Our method achieves the state-of-the-art performance, improving existing methods by 1.42, 2.44 and 1.29 in terms of BLEU-4, ROUGE-L and METEOR.

Retrieval-Augmented Generation for Code Summarization via Hybrid GNN

TL;DR

This work tackles automatic code summarization by bridging retrieval-based and generation-based approaches through a retrieval-augmented Hybrid Graph Neural Network (HGNN). HGNN constructs a retrieval-augmented static Code Property Graph and a structure-aware attention-based dynamic graph, enabling Hybrid Message Passing that captures both local and global code semantics. A retrieval-informed decoder and an attention-based LSTM generate succinct natural language summaries, with significant improvements on the CCSD and PCSD benchmarks, including BLEU-4, ROUGE-L, and METEOR improvements over prior methods. The approach introduces a challenging 95k+ function dataset (CCSD) and demonstrates strong generalization, human-judged quality, and qualitative case studies, highlighting its practical impact for code understanding and maintenance.

Abstract

Source code summarization aims to generate natural language summaries from structured code snippets for better understanding code functionalities. However, automatic code summarization is challenging due to the complexity of the source code and the language gap between the source code and natural language summaries. Most previous approaches either rely on retrieval-based (which can take advantage of similar examples seen from the retrieval database, but have low generalization performance) or generation-based methods (which have better generalization performance, but cannot take advantage of similar examples). This paper proposes a novel retrieval-augmented mechanism to combine the benefits of both worlds. Furthermore, to mitigate the limitation of Graph Neural Networks (GNNs) on capturing global graph structure information of source code, we propose a novel attention-based dynamic graph to complement the static graph representation of the source code, and design a hybrid message passing GNN for capturing both the local and global structural information. To evaluate the proposed approach, we release a new challenging benchmark, crawled from diversified large-scale open-source C projects (total 95k+ unique functions in the dataset). Our method achieves the state-of-the-art performance, improving existing methods by 1.42, 2.44 and 1.29 in terms of BLEU-4, ROUGE-L and METEOR.

Paper Structure

This paper contains 27 sections, 7 equations, 2 figures, 5 tables.

Table of Contents

  1. Introduction
  2. Hybrid GNN Framework
  3. Problem Formulation
  4. Retrieval-augmented Static Graph
  5. Graph Initialization
  6. Retrieval-based Augmentation
  7. Attention-based Dynamic Graph
  8. Hybrid GNN
  9. Decoder
  10. Experiments
  11. Setup
  12. Comparison with the baselines
  13. Ablation Study
  14. Human Evaluation
  15. Case Study
  16. ...and 12 more sections

Figures (2)

  • Figure 1: The overall architecture of the proposed HGNN framework.
  • Figure 2: An example of Code Property Graph (CPG).