Aprendizado de máquina multivisão aplicado à análise de correferência em um sistema de aprendizado sem fim

• Autor(a) principal: Mansano, Alex Fernandes
• Data de Publicação: 2018
• Tipo de documento: Dissertação
• Idioma: por
• Título da fonte: Repositório Institucional da UFSCAR
• Texto Completo: https://repositorio.ufscar.br/handle/ufscar/10227
• topic: Meta-heurística; Combinação de descritores; Redes neurais; Never-ending language learning; Meta-heuristics; Descriptor combination; Neural networks; CIENCIAS EXATAS E DA TERRA::CIENCIA DA COMPUTACAO::MATEMATICA DA COMPUTACAO; CIENCIAS EXATAS E DA TERRA::CIENCIA DA COMPUTACAO::METODOLOGIA E TECNICAS DA COMPUTACAO; CIENCIAS EXATAS E DA TERRA::CIENCIA DA COMPUTACAO::SISTEMAS DE COMPUTACAO; CIENCIAS EXATAS E DA TERRA::CIENCIA DA COMPUTACAO::TEORIA DA COMPUTACAO

NELL (Never-Ending Language Learning) is the first never-ending learning system presented in the litera ture. It has been modeled to create a knowledge base in an autonomous way, reading the web 24 hours per day, seven days per week. In this paradigm, all knowledge acquired is used to improve the learning performance. In this paradigm we face cases where the same object can be named in several ways. These cases as called as correferents, and has great importance for the never-ending learning process, as long as the knowledge about certain entity in a textual base may be distributed among its denominations.As such, the co-reference analysis has a crucial role in NELL’s learning paradigm. In this paper, we approach the combination of different feature vectors as an optimization task performed by meta-heuristic techniques and artificial neural networks, in order to maximize the separability of samples in the feature space, being the optimization process guided by the accuracy of Optimum Path Forest and variations of Siamese Networks in a validation set. The experiments showed the proposed methodology can obtain much better results when compared to the performance of individual feature extraction algorithms.