The Good, the Bad, and the Missing: Neural Code Generation for Machine Learning Tasks

Review Information Science & Library Science

A study on application programming interface recommendation: state-of-the-art techniques, challenges and future directions

Muhammad Sajid Nawaz et al.

Summary: This study conducts a systematic literature review and analysis of 35 primary studies to identify the challenges faced by developers, current state-of-the-art API recommendation techniques, and performance evaluation metrics. It presents outlines of a knowledge-based API recommendation system and provides a taxonomy and future research directions in the field of API recommendations.

LIBRARY HI TECH (2023)

添加到收藏夹

Article Computer Science, Software Engineering

Holistic Combination of Structural and Textual Code Information for Context Based API Recommendation

Chi Chen et al.

Summary: In this paper, a new API recommendation approach called APIRec-CST is proposed, which combines structural and textual code information using a deep learning model. The results show that this approach outperforms other methods in terms of accuracy and MRR, and textual code information is proven to be helpful in improving accuracy and MRR. Moreover, a user study demonstrates the practical utility of the proposed method.

IEEE TRANSACTIONS ON SOFTWARE ENGINEERING (2022)

添加到收藏夹

Article Computer Science, Software Engineering

Embedding API dependency graph for neural code generation

Chen Lyu et al.

Summary: The paper introduces a new method called ADG-Seq2Seq, which utilizes API dependency graph and embedding to generate code, showing significant improvements compared to existing methods. Extensive ablation tests demonstrate the effectiveness of ADG embedding.

EMPIRICAL SOFTWARE ENGINEERING (2021)

添加到收藏夹

Proceedings Paper Computer Science, Software Engineering

Empirical Study of Transformers for Source Code

Nadezhda Chirkova et al.

Summary: Initially developed for natural language processing, Transformers are now widely used for source code processing due to the format similarity. Recent works focus on developing Transformer modifications for capturing syntactic information in source code. This study empirically investigates the capabilities of Transformers to utilize syntactic information in different tasks, showing meaningful predictions based on syntax.

PROCEEDINGS OF THE 29TH ACM JOINT MEETING ON EUROPEAN SOFTWARE ENGINEERING CONFERENCE AND SYMPOSIUM ON THE FOUNDATIONS OF SOFTWARE ENGINEERING (ESEC/FSE '21) (2021)

添加到收藏夹

Proceedings Paper Computer Science, Information Systems

Self-Supervised Contrastive Learning for Code Retrieval and Summarization via Semantic-Preserving Transformations

Nghi D. Q. Bui et al.

Summary: Corder is a self-supervised contrastive learning framework designed for source code models, eliminating the need for labeled data in code retrieval and summarization tasks. By training the source code model to recognize similar and dissimilar code snippets through contrastive learning, Corder's pre-trained models have shown superior performance in code-to-code retrieval, text-to-code retrieval, and code-to-text summarization tasks compared to other baseline models.

SIGIR '21 - PROCEEDINGS OF THE 44TH INTERNATIONAL ACM SIGIR CONFERENCE ON RESEARCH AND DEVELOPMENT IN INFORMATION RETRIEVAL (2021)

添加到收藏夹

Proceedings Paper Computer Science, Software Engineering

Studying the Usage of Text-To-Text Transfer Transformer to Support Code-Related Tasks

Antonio Mastropaolo et al.

Summary: Deep learning techniques, particularly the T5 model, have gained attention in the software engineering community for their ability to support code-related tasks. By pre-training on a large dataset and fine-tuning on specialized data, T5 has shown improved performance in NLP tasks. This paper empirically investigates the performance of T5 in code-related tasks, demonstrating enhancements over previous DL-based solutions.

2021 IEEE/ACM 43RD INTERNATIONAL CONFERENCE ON SOFTWARE ENGINEERING (ICSE 2021) (2021)

添加到收藏夹

Article Computer Science, Information Systems