Conversa com Diana Sousa (LASIGE, DI-FCUL).
Successful biomedical relation extraction can provide evidence to researchers about possible unknown associations between entities, advancing our current knowledge about those entities and their inherent processes. Multiple relation extraction approaches have been proposed to identify relations between literature concepts, namely using neural networks algorithms. However, most approaches do not use external sources of knowledge as added resources, such as domain-specific ontologies. Moreover, dataset availability is still limited to train reliable models. In this seminar, I will present the advantages of integrating multiple ontologies into a BI-LSTM deep learning system. Additionally, how to create high-quality biomedical training corpora by allying distant supervision with crowdsourcing. The ultimate goal of the thesis is to use external semantic sources of knowledge along with the latest state-of-the-art language representations to improve the current performance of biomedical relation extraction both in English and non-English languages.
Short Bio: Diana Sousa received the B.S. degree in biochemistry and the M.Sc. degree in bioinformatics and computational biology from the Faculty of Sciences, University of Lisbon, Lisbon, Portugal, in 2017 and 2019, respectively, where she is currently pursuing the Ph.D. degree in informatics. Since 2016, she has been a Researcher with the LASIGE, Department of Informatics, Faculty of Sciences, University of Lisbon. Her research interest includes the development of text mining systems that aim at extracting biomedical information from biomedical literature, mainly focusing on relations between biomedical entities and on using deep learning approaches that resort to external sources of knowledge.
Transmissão em direto via Zoom.
