Explainable Assessment of Healthcare Articles with QA

Anonymous

Abstract: The healthcare domain suffers from the spread of poor quality articles on the Internet. While manual efforts exist, they are not sufficient to assess the amount of articles in circulation. The task can be automated as text classification, however explanations for the labels are necessary for the users. While current explainable systems tackle explanation generation as summarization, we propose a new approach based on Question-Answering that allows us to generate explanations for multiple criteria. We show that QA-based models are competitive with current state-of-the-art systems and complement summarization-based models for explainable quality assessment.