Automatic Extraction of Food-Gut Microbiota Relations in Scientific Publications

In recent years, the influence of nutrition on human health has become an important research field. It has shown the capacity to improve the recovery in ill patients by changing up the population of specific microbes. With the increasing amount of scientific publications, it’s necessary to develop new tools that help in the extraction of found relations in the healthcare community. In this ongoing work, we present a pipeline designed to automatically extract how aliments can affect the increase or decrease of gut microbes. For this workflow, we implemented a system that processes the given entities and uses public databases for the document compilation. Then, we used a series of state-of-the-art models, both for Named-Entity Recognition and for Relation Extraction, to locate the mention of the entities on the documents and to obtain the interactions detected between them. We also generated a small dataset for evaluating the models used from a set of papers labeled by experts.