Improving Bug Localization With Effective Contrastive Learning Representation

Abstract

Automated localization of buggy files can accelerate developers' efficiency of software maintenance, improving the quality of software products. State-of-the-art approaches for bug localization is based on neural networks, e.g., RNN or CNN, and can learn semantic feature from the given bug report. However, these simple neural architectures are difficult to learn the deep contextual feature from bug reports, which hurts the semantic mapping between bug reports and their corresponding buggy files. To resolve the above problem, in this paper we propose a bug localization approach that combines pre-trained language models and contrastive learning, namely CoLoc. Specifically, CoLoc first is pre-trained on a large-scale bug report corpus in an unsupervised way, to learn the deep contextual feature of each token in the bug report according to its context. Afterward, CoLoc is further pre-trained by a contrastive learning objective to learn the contrastive learning representations both of bug reports and buggy files. Contrastive learning can help CoLoc to learn the semantic differences between different bug reports and buggy files. To evaluate the effectiveness of CoLoc, we choose five baseline approaches and compare their performance on a public dataset. The experimental results show that CoLoc outperforms all baseline approaches and achieves new results for bug localization.

Insights of effectivity analysis of learning-based approaches towards software defect prediction

“…The study model formulates a syntax tree to represent the vectors linked with programs where the extraction of semantics is carried out by CNN model and retention of key features are carried out by Bi-LSTM. A unique form of deep learning approach known as contrastive learning is implemented by Luo et al [45] towards enhancing the identification of software defect in the form of bug. The study model initially performs pretraining on the corpus of bug reports using unsupervised scheme of learning followed by training with contrastive learning.…”

AttentiveBugLocator: A Bug Localization Model using Attention-based SemanticFeatures and Information Retrieval

“…However, the lexical mismatch remains a key issue in the effectiveness of these methods for bug localization. Very recent studies show that suitable representations of bug reports and source files using both deep learning and IR methods can improve the effectiveness of bug localization [16,7].…”

“…However, lexical mismatch remains a key issue to the effectiveness of these methods for bug localization. Meanwhile, representational learning has served to be the best approach to extract rich semantic and contextual features of bug reports and source code [38,16]. Notwithstanding, finding suitable representation to exploit various information -textual and statistical and stack traces -from bug reports, and code complexity and program structure in order to reduce the lexical mismatch remains a key challenge.…”