Can Plausibility Help to Support High Quality Content in Digital Libraries?

Presented herein is a novel approach to support high quality content in Digital Libraries by introducing the notion of Plausibility of new scientific papers when contrasted with prior knowledge. In particular, our work proposes a novel assessment of scientific papers to support the workload of reviewers. The proposed approach focus on a core component of a scientific paper: its claim. Our methodology exploits state of the art neural embedding representation of text and topic modeling on a Digital Library of scientific papers crawled from PubMed. As a proof of concept of the potential usefulness of the notion of Plausibility, we study and report experiments on documents with claims expressed as statistical associations. This type of claims is very often found in medicine, chemistry, biology, nutrition, etc. where the consumption of a drug, substance, product, etc., has an effect on some other type of entity such as a disease, another drug, sub-