Improving Long Covid-related Text Classification: A Novel End-to-end Domain-adaptive Paraphrasing Framework

This paper proposes a novel solution to the challenge of identifying relevant literature related to long COVID by employing machine learning techniques to classify long COVID literature. To overcome the scarcity of annotated data for machine learning, a strategy called medical paraphrasing is introduced to diversify the training data while maintaining the original content. Additionally, a Data-Reweighting-Based Multi-Level Optimization Framework for Domain Adaptive Paraphrasing is proposed, supported by a Meta-Weight-Network (MWN). This approach incorporates feedback from the downstream text classification model to influence the training of the paraphrasing model. The findings demonstrate that this method substantially improves the accuracy and efficiency of long COVID literature classification.