Multi-label classification for low-resource issue tickets with BERT

Nowadays, many companies, especially in IT, use project management tools to manage and track ongoing projects. Issue tickets within the tools can be assigned to any team member with details about the task. However, failing to label tickets with task categories affects progress monitoring. The research addresses the multi-label classification problem and the challenge of low-resource dataset. The dataset in this study is different from the large datasets usually used in current multi-label classification studies. Therefore, this research proposes embedding techniques to handle this situation. The methodology in this study applies several embedding techniques, including TF-IDF, Word2Vec, and BERT (WangchanBERTa and PhayaThaiBERT). The classification models, such as Logistic Regression, Support Vector Classifiers, Decision Trees, Neural Networks, and Deep Neural Networks, are experimented in this study to assess their performance. According to the experimental results, a combination of PhayaThaiBERT and Deep Neural Networks outperforms conventional methods. The combination yielded an F1 score of 0.769 on the test dataset.