Advancing COVID-19 Inquiry Responses Through Transfer Learning-Based Question Entailment Methodology

The COVID-19 pandemic underscores the urgency for efficient solutions addressing critical questions, given the prevalence of misinformation and ambiguous answers that necessitate a dependable automated system. Public Health Question Answering (QA) within Natural Language Processing (NLP) and Artificial Intelligence (AI) poses unique challenges, demanding both linguistic comprehension and domain expertise. Recent advancements in Deep Learning (DL) and Deep Neural Networks (DNN) have revolutionized QA systems, leading to improved performance. In this paper, we propose a novel question entailment approach named CoviBioBERT. It employs finely-tuned ensemble neural networks to enhance precision in addressing COVID-19-related queries. CoviBioBERT is an optimized QA system based on the BioBERT pre-trained model, customized for COVID-19 inquiries, and evaluated on a benchmark dataset of question-answer pairs. The system aims to provide dependable and accurate responses to contextually relevant COVID-19 queries. Experimental findings underscore the superiority of CoviBioBERT over foundational models, exhibiting a notably higher Exact Match (EM) score and ensuring more trustworthy responses to COVID-19-related inquiries. This paper makes a significant contribution by introducing a state-of-the-art QA system for retrieving COVID-19 information. It thereby equips the general public and healthcare practitioners with reliable insights to bolster global efforts against the pandemic.