Recognizing pharmacovigilance named entities in Brazilian Portuguese with CoreNLP

Abstract

Textual data sources may assist in the detection of adverse events not predicted for a particular drug. However, given the amount of information available in several sources, it is reasonable to adopt a computational approach to analyze these sources to search for adverse events. In this scenario, we created an extension of CoreNLP to process Brazilian Portuguese texts from pharma- covigilance area. We trained three natural language models: a Part-of-speech tagger, a parser and a Named Entity Recognizer. Preliminary results indicate success in generating a dependency tree for phrases in the pharmacovigilance area and in identifying pharmacovigilance named entities.