Modular Tree Network for Source Code Representation Learning
This addresses the need for better program analysis tools by improving source code representation learning, though it is incremental in leveraging structural information.
The paper tackles the problem of learning representations for source code by proposing a modular tree network (MTN) that dynamically composes neural units based on abstract syntax trees, achieving state-of-the-art performance in program classification and code clone detection.
Learning representation for source code is a foundation of many program analysis tasks. In recent years, neural networks have already shown success in this area, but most existing models did not make full use of the unique structural information of programs. Although abstract syntax tree-based neural models can handle the tree structure in the source code, they cannot capture the richness of different types of substructure in programs. In this paper, we propose a modular tree network (MTN) which dynamically composes different neural network units into tree structures based on the input abstract syntax tree. Different from previous tree-structural neural network models, MTN can capture the semantic differences between types of ASTsubstructures. We evaluate our model on two tasks: program classification and code clone detection. Ourmodel achieves the best performance compared with state-of-the-art approaches in both tasks, showing the advantage of leveraging more elaborate structure information of the source code.