Aayush Panchal

Master's

Software Engineering

A BERT

- LogBERT is an advanced, automated log quality assessment tool for software systems that uses transformer models and deep learning techniques to analyse the static text of log messages - Implemented LogBERT architecture with pre-trained BERT models and custom neural networks, achieving a significant 23% increase in loglevel - Demonstrated LogBERT's effectiveness with average AUC scores of 0.94 and 0.98 for multi-level and binary classification tasks, respectively - Analyzed linguistic structure using part-of-speech tags and static text features, achieving perfect precision and recall balance with an F1 score - Addressed dataset labeling and preprocessing challenges, proposing solutions to enhance model accuracy, such as character length consideration