TransformCode: A Contrastive Learning Framework for Code Embedding via Subtree Transformation

Keyphrases

• Abstract Syntax Tree 33%
• Agnostic 33%
• Amount of Training 33%
• Artificial Intelligence 33%
• Artificial Intelligence for Software Engineering 33%
• Clone Classification 33%
• Code Clone Detection 33%
• Code Embedding 100%
• Code Representation 33%
• Code Snippets 33%
• Code2vec 33%
• Computing Resources 33%
• Contrastive Learning 100%
• Data Augmentation 33%
• Downstream Task 33%
• Encoder 66%
• Encoder Parameters 33%
• Labeled Data 33%
• Language-agnostic 33%
• Large Animal Model 33%
• Learning Framework 100%
• Learning Task 33%
• Parameter Dependence 33%
• Programming Languages 66%
• Semantic Transformation 33%
• Software Development Efficiency 33%
• Software Engineering 33%
• Specific Labels 33%
• State-of-the-art Techniques 33%
• Supervised Learning 33%
• Syntactic Transformations 33%
• Token 33%
• Training Data 33%
• Training Model 66%
• Transfer Learning 33%
• Tree Transformation 33%
• Unsupervised Learning 33%

Computer Science

• Abstract Syntax Tree 33%
• Artificial Intelligence 66%
• Clone Detection 33%
• Computing Resource 33%
• Contrastive Learning 100%
• Data Augmentation 33%
• Learning Framework 100%
• Programming Language 66%
• Software Development 33%
• Software Engineering 66%
• Supervised Learning 33%
• Syntactics 33%
• Training Data 33%
• Transfer Learning 33%
• Unsupervised Learning 33%