An Empirical Study of Source Code Detection Using Image Classification

The detection of programming language for a source code file has achieved high accuracy using the machine learning techniques. On the other hand, for a piece of software (called snippet), the detection of programming language is required to append tags automatically in a question and answer site such as Stack Overflow. However, the detection of programming language for a snippet is still a challenge since snippets is not a complete source code. Usually, experienced developers can detect the language of such snippet at a glance. It is considered that such a task that a human being easily solves can be solved by the image classification method using deep learning technique. Therefore, we propose a programming language detection method using a deep learning based image classification method. By using the data from actual Q&A site, we evaluate our proposed model. The results of experiment demonstrate that we can successfully detect the correct programming language for snippets with over 90% accuracy.