Speakers
Description
A medication package insert is a legal healthcare document with important information about medications. In Brazil, the National Health Surveillance Agency (ANVISA) requires two versions of the package insert: one for patients and another one for healthcare professionals. In this study, we manually evaluated the performance of an automatic frame annotator on a corpus consisting of 100 sentences targeting patients and 100 sentences targeting healthcare professionals. The aim was to evaluate whether the parser’s output evidenced correct assignment of semantic frames and their frame elements (FEs) in each input sentence and to what extent human post-annotation would be necessary to improve the output. Text target audience was defined as a variable potentially impacting frame detection given differences in the language of package inserts. Overall, the findings demonstrate the efficacy of the automated annotation process, revealing challenges that have to do with the same form being capable of classification under different categories and/or frames. Few differences were found when comparing sentences for the two distinct audiences targeted in the texts.
