Semi-supervised automatic semantic role labeling for brazilian portuguese

• Grant number: 10/04647-0
• Support Opportunities: Scholarships in Brazil - Master
• Effective date (Start): August 01, 2011
• Effective date (End): December 31, 2012
• Field of knowledge: Physical Sciences and Mathematics - Computer Science - Computer Systems
• Principal Investigator: João Luís Garcia Rosa
• Grantee: Fernando Emilio Alva Manchego
• Host Institution: Instituto de Ciências Matemáticas e de Computação (ICMC). Universidade de São Paulo (USP). São Carlos , SP, Brazil
• Associated scholarship(s): 11/22500-0 - Adapting unsupervised semantic role labeling methods for Brazilian Portuguese, BE.EP.MS

Abstract

Semantic Role Labeling (SRL) is a natural language processing (NLP) task which, in the last years, has been highly researched because it allows an analysis of the meaning of sentences through detecting the events that are being described in them, the same as the participants involved, which is essential for computers to effectively understand the information coded in text. However, most the research developed has been done for english text, considering the grammatical and semantic characteristics of that language, which doesn't allow that those products and results be directly transported for others such as portuguese. Supervised learning methods are used nowadays for automatic SRL, but for a successful learning, big corpora of annotated sentences are required. For brazilian portuguese it's being developed the PropBank.Br, which provides an annotated corpus, small but useful, for this task. Therefore, a method capable of extracting relevant information from those "few" labeled sentences and also abundant available unannotated data will be used (semi-supervised learning). With this objective, a classifier will be trained using the PropBank.Br labels for labeling the corpus Bosque (section CETENFolha) from the Floresta Sintá(c)tica, using the self-training algorithm and the maximum entropy models as basic classifier. Precision, recall and F1 will be calculated to evaluate the labeler's performance.At the end of the research, it is expected to make available a tool for automatic SRL for brazilian portuguese, which could benefit different areas of NLP for Portuguese. (AU)

Academic Publications

(References retrieved automatically from State of São Paulo Research Institutions)

MANCHEGO, Fernando Emilio Alva. Automatic semi-supervised semantic role labeling for Brazilian Portuguese. 2013. Master's Dissertation - Universidade de São Paulo (USP). Instituto de Ciências Matemáticas e de Computação (ICMC/SB) São Carlos.