Student Program Classification Using Gated Graph Attention Neural Network

Abstract

The online programing services, such as Github, TopCoder, and EduCoder, have promoted a lot of social interactions among the service users. However, the existing social interactions is rather limited and inefficient due to the rapid increasing of source-code repositories, which is difficult to explore manually. The emergence of source-code mining provides a promising way to analyze those source codes, so that those source codes can be relatively easy to understand and share among those service users. Among all the source-code mining attempts, program classification lays a foundation for various tasks related to source-code understanding, because it is impossible for a machine to understand a computer program if it cannot classify the program correctly. Although numerous machine learning models, such as the Natural Language Processing (NLP) based models and the Abstract Syntax Tree (AST) based models, have been proposed to classify computer programs based on their corresponding source codes, the existing works cannot fully characterize the source codes from the perspective of both the syntax and semantic information. To address this problem, we proposed a Graph Neural Network (GNN) based model, which integrates data flow and function call information to the AST, and applies an improved GNN model to the integrated graph, so as to achieve the state-of-art program classification accuracy. The experiment results have shown that the proposed work can classify programs with accuracy over 97%.

GRAPHSPY: Fused Program Semantic Embedding through Graph Neural Networks for Memory Efficiency

“…There are also several works learning from program representation to conduct program repair (Wang, Su, and Singh 2018;Wang and Su 2020), bug detection (Dinella et al 2020), program classification (Lu et al 2019), cache replacement (Li and Gu 2020;Liu et al 2020), heap memory wise program verification (Li et al 2016), and code similarity detection (Yu et al 2020). Our work is the first to apply GNNs for dead store detection.…”

“…A CFG represents intra-procedure program structure semantic, so we attempt to embed a CFG into a feature map. First, we apply the word2vec (Mikolov et al 2013) model to encode every token in an instruction inside a basic block. Then we average out token embeddings as an embedding of an instruction, and finally derive a whole embedding of a basic block by averaging the embeddings of instructions.…”

“…The basic building blocks of a procedure are the so-called basic blocks. By tokenizing a basic block into a set of words, we use word2vec (Mikolov et al 2013) to embed each word and average them as an embedding for representing a basic block. Viewing every basic block as a node, an intraprocedural syntactic structure, i.e.…”

Mapping Modern JVM Language Code to Analysis-friendly Graphs: A Pilot Study with Kotlin

“…Allamanis et al [3] mapped program to graphs consisting of ASTs together with control-flow edges, data-flow edges, and a hand-crafted set of additional typed edges. Lu M et al [4] proposed a program graph method named FDA, which integrates the AST, function call graph, and data-flow graph to characterize syntax and semantic information. T. T. Nguyen et al [5] proposed a program graph method named Groum, where nodes represent actions and control points, and edges represent control and data flow dependencies between nodes.…”

Text Summarization Method Based on Gated Attention Graph Neural Network

“…However, this requires a large amount of time to be reserved for data processing problems. To address this problem, a text summarization model combining Gated Graph Neural Networks (GGNN) [ 22 , 23 , 24 , 25 ] with attention is proposed, and the model has better performance than most baseline models. To improve the robustness of the graph structure, refs.…”