• JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
 
  Bookmark and Share
 
 
Master's Dissertation
DOI
https://doi.org/10.11606/D.3.2020.tde-14012021-124730
Document
Author
Full name
Gabriela Souza de Melo
E-mail
Institute/School/College
Knowledge Area
Date of Defense
Published
São Paulo, 2020
Supervisor
Committee
Cozman, Fabio Gagliardi (President)
Finger, Marcelo
Lopes, Marcos Fernando
Title in English
Winograd schemas in portuguese.
Abstract in English
The Winograd Schema Challenge has become a common benchmark for question answering and natural language processing, having been originally developed in the English language. In order to stimulate the development of Natural Language Processing in Portuguese, we have developed a set of Winograd Schemas in Portuguese. We have also adapted solutions proposed for the English-based version of the challenge so as to have an initial model for usage in its Portuguese-based version. To do so, we created a language model for Portuguese based on a set of Wikipedia documents. In order to evaluate the impact of the increase in the language model capacity in the results for the Portuguese challenge, we tested training of the model with the usage of a myriad of techniques that have previously led to state-of-the-art results for English-based tasks.
Title in Portuguese
Esquemas de Wwnograd em português.
Keywords in Portuguese
Análise de texto
Inteligência artificial
Linguagem natural
Sintaxe e semântica da linguagem natural
Abstract in Portuguese
O Desafio de Winograd se tornou uma referência em tarefas de resposta textual automatizada e processamento de linguagem natural, tendo sido desenvolvido inicialmente na língua inglesa. De forma a estimular o desenvolvimento do campo de Processamento de Linguagem Natural em português, desenvolvemos um conjunto de Esquemas de Winograd em português. Também adaptamos soluções propostas para a versão do desafio baseada em inglês, de forma a disponibilizarmos um modelo inicial para ser utilizado na versão do desafio baseada em português. Para fazê-lo, criamos um modelo de linguagem treinado sobre um conjunto de documentos da Wikipedia. De forma a avaliar o impacto da variação da capacidade do modelo de linguagem nos resultados para o desafio em português, nós testamos o treinamento do modelo utilizando-se de diversas técnicas que anteriormente levaram a resultados estado-da-arte para tarefas na língua inglesa.
 
WARNING - Viewing this document is conditioned on your acceptance of the following terms of use:
This document is only for private use for research and teaching activities. Reproduction for commercial use is forbidden. This rights cover the whole data about this document as well as its contents. Any uses or copies of this document in whole or in part must include the author's name.
Publishing Date
2021-01-19
 
WARNING: Learn what derived works are clicking here.
All rights of the thesis/dissertation are from the authors
CeTI-SC/STI
Digital Library of Theses and Dissertations of USP. Copyright © 2001-2021. All rights reserved.