Defesa de Dissertação: Bárbara Stéphanie Neves Oliveira
Data da publicação: 23 de agosto de 2024 Categoria: Defesas de Dissertação, Notícias
Título: Exploring the Dependence of Text Embeddings on the Target of NLP Tasks
Data: 30/08/2024
Horário: 14h00
Local: Online
Resumo:
In many human languages, linguistic units represent text structure. Text embeddings, a core component of Natural Language Processing (NLP), transform these units into dense vector representations that encapsulate semantic and syntactic nuances. However, evaluating these embeddings for specific tasks remains a complex process involving intrinsic and extrinsic approaches. This work explores the dependencies and correlations between text embeddings and their performance in different NLP tasks. It highlights challenges such as the absence of a universal embedding method and the potential for overfitting in advanced models. Through two key contributions and research outcomes, including the HELD framework for entity-label disambiguation and a study on embedding dependence, the research offers insights into the nuances of embedding evaluation.
Banca examinadora:
- Prof. Dr. José Antonio Fernandes de Macêdo (MDCC/UFC) – Orientador
- Profa. Dra. Ticiana Linhares Coelho da Silva (UFC) – Coorientadora
- Prof. Dr. César Lincoln Cavalcante Mattos (MDCC/UFC)
- Profa. Dra. Lívia Almada Cruz (UFC)
