NMT-Based Code Generation for Coding Assistance with Natural Language

• Item type: Trans(1)
• 公開日: 2022-05-20
• タイトル: NMT-Based Code Generation for Coding Assistance with Natural Language
• タイトル / 言語: en
• タイトル: NMT-Based Code Generation for Coding Assistance with Natural Language
• 言語: eng
• キーワード / 主題Scheme: Other
• キーワード / 主題: [通常論文] neural machine translation, code generation, back-translation
• 資源タイプ / 資源タイプ識別子: http://purl.org/coar/resource_type/c_6501
• 資源タイプ: journal article
• 著者所属: Division of Mathematical and Physical Sciences, Graduate School of Science, Japan Women's University／Presently with NTT Software Innovation Center
• 著者所属: Division of Mathematical and Physical Sciences, Graduate School of Science, Japan Women's University
• 著者所属: Division of Mathematical and Physical Sciences, Graduate School of Science, Japan Women's University
• 著者所属: Department of Mathematics, Physics, and Computer Science, Japan Women's University
• 著者所属(英): en
• 著者所属(英): Division of Mathematical and Physical Sciences, Graduate School of Science, Japan Women's University / Presently with NTT Software Innovation Center
• 著者所属(英): en
• 著者所属(英): Division of Mathematical and Physical Sciences, Graduate School of Science, Japan Women's University
• 著者所属(英): en
• 著者所属(英): Division of Mathematical and Physical Sciences, Graduate School of Science, Japan Women's University
• 著者所属(英): en
• 著者所属(英): Department of Mathematics, Physics, and Computer Science, Japan Women's University
• 著者名: Yuka, Akinobu; Teruno, Kajiura; Momoka, Obara; Kimio, Kuramitsu
• 著者名(英): Yuka, Akinobu; Teruno, Kajiura; Momoka, Obara; Kimio, Kuramitsu
• 論文抄録 / 内容記述タイプ: Other
• 論文抄録 / 内容記述: This paper proposes an attempt to realize coding assistance that generates Python code from natural language descriptions using neural machine translation. Although coding assistance with deep learning has recently become a major concern, few applications have used neural machine translation models. One of the major barriers is the shortage of a parallel corpus of natural language descriptions and source code. To overcome the shortage of parallel corpora, we propose a method for synthesizing parallel corpora that utilizes the formal nature of programming languages. We aim to establish a new method using an abstract syntax tree (AST) and a corpus of code fragments. Using the proposed synthesis method, we successfully constructed tens of thousands of parallel corpora and trained PyNMT models to generate Python code from Japanese input sentences. The trained PyNMT models successfully predicted the Python code from user input sentences with an accuracy of 28%. In this study, we propose a synthetic method for a parallel corpus and summarize the results of the evaluation experiments conducted on PyNMT models.; ------------------------------; This is a preprint of an article intended for publication Journal of; Information Processing(JIP). This preprint should not be cited. This; article should be cited as: Journal of Information Processing Vol.30(2022) (online); ------------------------------
• 論文抄録(英) / 内容記述タイプ: Other
• 論文抄録(英) / 内容記述: This paper proposes an attempt to realize coding assistance that generates Python code from natural language descriptions using neural machine translation. Although coding assistance with deep learning has recently become a major concern, few applications have used neural machine translation models. One of the major barriers is the shortage of a parallel corpus of natural language descriptions and source code. To overcome the shortage of parallel corpora, we propose a method for synthesizing parallel corpora that utilizes the formal nature of programming languages. We aim to establish a new method using an abstract syntax tree (AST) and a corpus of code fragments. Using the proposed synthesis method, we successfully constructed tens of thousands of parallel corpora and trained PyNMT models to generate Python code from Japanese input sentences. The trained PyNMT models successfully predicted the Python code from user input sentences with an accuracy of 28%. In this study, we propose a synthetic method for a parallel corpus and summarize the results of the evaluation experiments conducted on PyNMT models.; ------------------------------; This is a preprint of an article intended for publication Journal of; Information Processing(JIP). This preprint should not be cited. This; article should be cited as: Journal of Information Processing Vol.30(2022) (online); ------------------------------
• 書誌レコードID / 収録物識別子タイプ: NCID
• 書誌レコードID / 収録物識別子: AA11464814
• 書誌情報: 情報処理学会論文誌プログラミング（PRO）; 巻 15, 号 2, 発行日 2022-05-20
• ISSN / 収録物識別子タイプ: ISSN
• ISSN / 収録物識別子: 1882-7802
• 出版者 / 言語: ja
• 出版者: 情報処理学会