Online News Sentiment Classification Using DistilBERT

The ability of pre-trained BERT model to achieve outstanding performances on many Natural Language Processing (NLP) tasks has attracted the attention researchers in recent times. However, huge computational and memory requirements have hampered its widespread deployment devices with limited resources. concept knowledge distillation shown produce smaller faster distilled models less trainable parameters intended for resource-constrained environments. can be fine-tuned great performance a wider range tasks, such as sentiment classification. This paper evaluates DistilBERT other pre-canned text classifiers Covid-19 online news binary classification dataset. analysis shows that despite having fewer than BERT-based model, achieved an accuracy 0.94 validation set after only two training epochs. also highlights usefulness ktrain library facilitating building, training, application state-of-the-art Machine Learning Deep models.