通过上下文数据丰富源代码，用于代码完成模型的实证研究（英文版）

英文标题：Enriching Source Code with Contextual Data for Code Completion Models: An Empirical Study中文摘要：本研究旨在探索是否通过使用上下文数据使代码易于理解会提高预训练代码语言模型完成代码补全任务的性能，并建议在训练、微调或选择此类模型时做出适当的设计选择，以改善自动补全的实用性和准确性。英文摘要：Transformer-based pre-trained models have recently achieved great results insolving many software engineering tasks including automatic code completionwhich is a staple in a developer's toolkit. While many have striven to improvethe code-understanding abilities of such models, the opposite -- making th