Abstract Text classification is a widely studied problem and has broad applications. In many real-world problems, the number of texts for training models limited, which renders these prone to overfitting. To address this problem, we propose SSL-Reg, data-dependent regularization approach based on self-supervised learning (SSL). SSL (Devlin et al., 2019a) an unsupervised that defines auxiliary t...

Problems and strategies for text classification have already been known a long time. Theyre widely utilised by companies like Google Yahoo email spam screening, sentiment analysis of Twitter data, automatic news categories in alerts. Were still working on getting the findings to be as accurate possible. When dealing with large amounts however, models performance accuracy become difficulty. The ...