Analysis of natural disasters in data from news

• Ano de defesa: 2024
• Autor(a) principal: Garcia, Klaifer [UNIFESP]
• Orientador(a): Berton, Lilian [UNIFESP]
• Tipo de documento: Tese
• Tipo de acesso: Acesso aberto
• dARK ID: ark:/48912/001300002ssws
• Idioma: eng
• Instituição de defesa: Universidade Federal de São Paulo
• Palavras-chave em Português: Natural Language Processing; Automatic Text Summarization; Event Detection; Automatic Text Classification; Machine Learning
• Link de acesso: https://hdl.handle.net/11600/72677

Natural disasters have been occurring with increasing frequency as a result of human activity on the environment, causing significant damage to society. Minimizing these losses depends on the development of protection policies, which need to be supported by accurate information about the events. However, collecting information on disasters presents several challenges, such as insufficient manpower to document every detail of the event and the unpredictability of the events, making it difficult to capture the initial moments after a disaster. In light of these challenges, this work developed methodologies to utilize news data as an alternative source of information on disasters. Specifically, techniques for document filtering, event detection, and automatic summarization were proposed and optimized to achieve better results in this domain, with a particular focus on improving applications in Portuguese, as there is a shortage of research in this language. The main contributions of this work are: 1) a complete framework for building knowledge bases from news articles, 2) new Portuguese datasets for several Natural Language Processing (NLP) tasks, 3) a novel method to produce more accurate summaries based on siamese networks, 4) an evaluation of the latest text classification techniques for application in Portuguese, and 5) a systematic literature review on event detection in news. This work provides contributions to various NLP tasks, with a special emphasis on addressing and developing solutions for the Portuguese language.