Graph neural networks for relationship extraction from biomedical literature
Graph Neural Networks (GNNs) have recently emerged as a powerful tool for mining relationships in unstructured data. In the biomedical domain, where text often contains intricate relationships between entities, traditional methods fall short in capturing the complexity of these interactions. This st...
| Autor: | |
|---|---|
| Formato: | tesis de maestría |
| Fecha de publicación: | 2024 |
| País: | España |
| Recursos: | Universitat Politècnica de Catalunya (UPC) |
| Repositorio: | UPCommons. Portal del coneixement obert de la UPC |
| Idioma: | inglés |
| OAI Identifier: | oai:upcommons.upc.edu:2117/408502 |
| Acesso em linha: | https://hdl.handle.net/2117/408502 |
| Access Level: | acceso abierto |
| Palavra-chave: | Natural language processing (Computer science) Neural networks (Computer science) Natural Language Processing Relationship Extraction Graph Neural Networks BERT Large Language Models Processament del Llenguatge Natural Extracció de Relacions Xarxes Neuronals de Grafs Models de Llenguatge Grans Tractament del llenguatge natural (Informàtica) Xarxes neuronals (Informàtica) Àrees temàtiques de la UPC::Informàtica::Sistemes d'informació |
| Resumo: | Graph Neural Networks (GNNs) have recently emerged as a powerful tool for mining relationships in unstructured data. In the biomedical domain, where text often contains intricate relationships between entities, traditional methods fall short in capturing the complexity of these interactions. This study explores a novel approach leveraging GNNs for relationship extraction in biomedical text mining, aiming to enhance the extraction of relationships between entities in annotated biomedical literature. By constructing a word graph from annotated text and applying GNNs, our method aims to outperform existing state-of-the-art techniques, which predominantly combine Bidirectional Encoder Representations from Transformers (BERT) with GNNs in a disparate fashion. Our results identify some challenges associated with formulating relationship extraction as a GNN Edge Classification task. Nevertheless, this paper underscores the potential of integrating GNNs with Large Language Models (LLMs) for Biomedical Natural Language Processing (BioNLP) and sets a foundation for future research that could extend these methods to broader applications in the field, potentially contributing to advancements in biomedical research for pharmacological drug discovery and healthcare applications. |
|---|