# A Survey of Deep Code Search

**Authors:** Yutao Xie, Jiayi Lin, Hande Dong, Lei Zhang, Zhonghai Wu  
**arXiv:** [2305.05959](https://arxiv.org/abs/2305.05959)  
**Date:** 2023-05-10  
**Summary page:** https://www.summarizepaper.com/en/arxiv-id/2305.05959v1/

## Key points

- Survey on Deep Code Search by Yutao Xie, Jiayi Lin, Hande Dong, Lei Zhang, and Zhonghai Wu
- Importance of capturing semantic similarity between queries and code snippets for effective information retrieval
- Deep learning as a powerful tool for extracting complex semantic information in code search
- Application of deep learning methods like graph neural networks and pretraining models to improve accuracy and efficiency
- Overview of deep learning-based code search frameworks mapping queries and codes to vectors for similarity measurement
- Proposed taxonomy illustrating the three-step process in deep learning-based code search methodology: query semantics modeling, code semantics modeling, and matching modeling involving model training
- Outlined potential avenues for future research in the field of deep learning-based code search

## Summary

The survey titled "A Survey of Deep Code Search" by authors Yutao Xie, Jiayi Lin, Hande Dong, Lei Zhang, and Zhonghai Wu delves into the realm of code intelligence techniques with a specific focus on code search. The repetitive and predictable nature of code writing has prompted the development of various methodologies to enhance code search capabilities. This survey emphasizes the importance of capturing semantic similarity between queries and code snippets to retrieve relevant information effectively. Deep learning has emerged as a powerful tool in extracting complex semantic information, leading to significant advancements in the field of code search. Various deep learning methods such as graph neural networks and pretraining models have been successfully applied to improve the accuracy and efficiency of code retrieval processes. As a result, deep learning has become the predominant paradigm for code search. The survey provides a comprehensive overview of deep learning-based code search frameworks that map queries and codes to vectors, enabling the measurement of their similarity. A new taxonomy is proposed to illustrate the state-of-the-art deep learning-based code search methodology in a three-step process: query semantics modeling, code semantics modeling, and matching modeling involving deep learning model training. Furthermore, the survey outlines potential avenues for future research in this promising field, highlighting opportunities for further exploration and innovation. By synthesizing existing knowledge and presenting a roadmap for future developments, this survey serves as a valuable resource for researchers and practitioners seeking to leverage deep learning techniques for enhanced code search capabilities.