Multi-labelled Topic Classification of Research Articles Using Machine Learning

Multi-label text classification is a type of classification where a single instance can be assigned with more than one class. It is much more difficult than single label classification as many labels tends to be correlated to one another. As an important task with comprehensive applications multi-label classification has been applied to many fields. One such field is topic classification. In this paper, we are going to classify research paper abstracts into different classes using various machine learning, deep learning, and attention-based models. We are going to use well-known metrics like accuracy, micro-precision, micro-recall, and micro-F1 score to compare these techniques. The results show that attention based models like robustly optimized bidirectional encoder representations from transformers model (RoBERTa) gives the best accuracy out of all the models.