Medical papers are being widely published currently, especially after the Coronavirus disease (COVID-19) pandemic. The time required to manually summarize medical papers can be decreased by applying text summarization approaches. It is now common practice to overcome medical text summarization challenges using pre-trained models such as the Bidirectional Encoder Representations from Transformers (BERT)-base model. This paper presents a new system for summarizing medical papers based on deep learning techniques. In this system, we combine the-Statistic (CHI-square) feature selection technique with a token classification such as Part-of-Speech (POS) tagging and use the feature selection output as input to the pre-training BERT-base model, then apply clustering algorithms for the sentence selection process. Our main contribution is that our model obtained high speed and accuracy compared
Link: Speed Up The Deep Bidirectional Transformers With Feature Selection For Summarizing Medical Papers
