Incm & Nova Sbe Pbl project - retirement law query system: performance analysis & evaluation of bert for extractive question-answering of portuguese retirement law

• Autor(a) principal: Araújo, Ricardo Luís Ferreira Morais de
• Data de Publicação: 2021
• Tipo de documento: Dissertação
• Idioma: eng
• Título da fonte: Repositório Científico de Acesso Aberto de Portugal (Repositórios Cientìficos)
• Texto Completo: http://hdl.handle.net/10362/153749
• topic: Data science; Business analytics; Business and data analytics; Nlp; Bert; Extractive qa; Semantic search; Domínio/Área Científica::Ciências Sociais::Economia e Gestão

This study presents a performance evaluation of the model BERT for an Extractive Question Answering task applied to the domain of Portuguese Retirement Law. By using multiple standard metrics in information retrieval solutions, the study measures how the Portuguese BERT performs as a semantic search solution on Portuguese Retirement Law by comparing its answers with a testing set composed of 180 questions and answers provided by INCM. The study revealed that the solution showed positive results. A form was also distributed to INCM, which indicated positive feedback when comparing it with the current solution implemented. This project showed that the Portuguese BERT is a very promising tool that can be used for Extractive Question Answering and for semantic search on various other domains.