Comparing Code Similarity Using Information Retrieval Techniques and Deep Learning
-
- YOKOI Kazuki
- Osaka University
-
- CHOI Eunjong
- Kyoto Institute of Technology
-
- YOSHIDA Norihiro
- Ritsumeikan University
-
- MATSUSHITA Makoto
- Osaka University
-
- INOUE Katsuro
- Nanzan University
Bibliographic Information
- Other Title
-
- 情報検索技術と深層学習を用いたコード片類似性判定法の比較調査
Abstract
Measuring code similarity is a fundamental technique in software engineering. It is challenging to measure code similarity not only syntactical but also functional. Previous research proposed measuring functional similarity using information retrieval (IR) techniques. Recently, measurement methods using deep learning have also been proposed. They have different pros and cons in terms of accuracy and calculation time. In this paper, we compare the combination of IR and deep learning for code similarity. As a result, the combination of LSI (Latent Semantic Indexing), a sort of IR technique, and a deep learning model showed the highest accuracy and fastest calculation time.
Journal
-
- 電子情報通信学会論文誌D 情報・システム
-
電子情報通信学会論文誌D 情報・システム J106-D (4), 231-243, 2023-04-01
The Institute of Electronics, Information and Communication Engineers
- Tweet
Details 詳細情報について
-
- CRID
- 1390858529773389440
-
- ISSN
- 18810225
- 18804535
-
- HANDLE
- 11094/93106
-
- Text Lang
- ja
-
- Data Source
-
- JaLC
- IRDB
- KAKEN
-
- Abstract License Flag
- Disallowed