Using a noun phrases micro reader to aid in the process of never ending learning

• Grant number: 16/16536-5
• Support Opportunities: Scholarships in Brazil - Scientific Initiation
• Start date: January 01, 2017
• End date: December 31, 2017
• Field of knowledge: Physical Sciences and Mathematics - Computer Science - Computing Methodologies and Techniques
• Principal Investigator: Estevam Rafael Hruschka Júnior
• Grantee: Luís Felipe Franco Candêo Tomazini
• Host Institution: Centro de Ciências Exatas e de Tecnologia (CCET). Universidade Federal de São Carlos (UFSCAR). São Carlos , SP, Brazil

Abstract

This project consists on investigating a way to use a micro reader of noun phrases to aid in the extraction of facts from the web in a never-ending learning system, in this case NELL - http://rtw.ml.cmu.edu/rtw/. The intention is to investigate and implement methods that allow the extraction of rules based on noun phrases, in the process of machine learning, from sentences. The aim is also to study how to use NELL's current macro reader in cooperation with the new micro reader. In order to achieve these goals, this project consists on developing a computational program capable of extracting knowledge through the micro reading based on noun phrases in Portuguese using the Conditional Random Fields - CRFs technique. The steps will be as follows: research and learning about previous related projects; creation of a baseline in English to be used as a support to the evaluation of the proposed methodology; investigation and incorporation of a CRF model to the baseline in English; development of a baseline in Portuguese, which should have equal or better performance than the baseline in English; and investigation and implementation of a different CRF model for the baseline in English, as there are structural differences between the two languages. There will be two distinct result analyses. The first one will be related to the characteristics of the software to be developed to allow the use of the new micro reader. The second one will be based on the results of the experiments from the integration of the never-ending learning system with the micro reader of noun phrases in Portuguese. Thus, NELL will run without the new micro reader and then with the micro reader. (AU)