On the contribution of specific entity detection and comparative construction to automatic spin detection in biomedical scientific publications

In this article we address the problem of providing automatized aid for the detection of misrepresentation (spin) of research results in scientific publications from the biomedical domain. For identifying automatically inadequate claims in medical articles, i.e. claims that state the beneficial effect of the experimental treatment to be greater than it is actually proven by the research results, we propose a Natural Language Processing (NLP) approach. We first make a review of related works and an NLP analysis of the problem; then we present our first results obtained on the type of publications most likely amenable to automatic processing: articles which present results of Randomized Controlled Trials (RCTs), i.e. comparisons done by applying the experimental or standard treatment on different registered patient groups. Our results concern the identification of specific entities necessarily present in an RCT description (here outcomes and patient groups), obtained with basic methods (local grammars) on a corpus extracted from the PubMed open archive. Then we describe our findings on the support we could gain by identifying comparative constructions and their relationship to the identified entities as preliminary step for deploying sentiment analysis as one of the constituent functionalities of our automatic spin detection algorithm