Por Diana Sousa (orientador: Francisco Couto).

Relation Extraction is a crucial process to deal with the amount of text published daily, for example, to find missing associations in a database. Relation Extraction is a text mining task for which the state-of-the-art approaches use bidirectional encoders, namely, BERT. However, most systems lack external knowledge injection, which is more problematic in the biomedical area given the widespread usage and high quality of biomedical ontologies. This knowledge can propel these systems forward by aiding them in predicting more explainable biomedical associations. With this in mind, in this seminar, I will present K-RET, a novel, knowledgeable biomedical relation extraction system that is, for the first time, able to handle different associations, integrate knowledge from multiple sources, define where to apply it, and deal with multi-token entities. K-RET was tested on three open-access corpora (DDI, BC5CDR, and PGR) integrated with four biomedical ontologies handling different entities. K-RET improved state-of-the-art results by 2.68% on average, with the DDI yielding the most significant boost in performance, from 79.60% to 87.88% in F-measure.