Clause identification using entropy guided transformation learning
Research output: Contributions to collected editions/works › Article in conference proceedings › Research › peer-review
Standard
2009 Seventh Brazilian Symposium in Information and Human Language Technology, STIL 2009: 8 - 11 September 2009, São Carlos, São Paulo, Brazil; Proceedings . Piscataway: IEEE - Institute of Electrical and Electronics Engineers Inc., 2009. p. 117-124 5532445.
Research output: Contributions to collected editions/works › Article in conference proceedings › Research › peer-review
Harvard
APA
Vancouver
Bibtex
}
RIS
TY - CHAP
T1 - Clause identification using entropy guided transformation learning
AU - Fernandes, Eraldo R.
AU - Pires, Bernardo A.
AU - dos Santos, Cícero N.
AU - Milidiú, Ruy L.
N1 - Conference code: 7
PY - 2009
Y1 - 2009
N2 - Entropy Guided Transformation Learning (ETL) is a machine learning strategy that extends Transformation Based Learning by providing automatic template generation. In this work, we propose an ETL approach to the clause identification task. We use the English language corpus of the CoNLL'2001 shared task. The achieved performance is not competitive yet, since the F β=1 of the ETL based system is 80:55, whereas the state-of-the-art system performance is 85:03. Nevertheless, our modeling strategy is very simple, when compared to the state-of-the-art approaches. These first findings indicate that the ETL approach is a promising one for this task. One can enhance its performance by incorporating problem specific knowledge. Additional features can be easily introduced in the ETL model.
AB - Entropy Guided Transformation Learning (ETL) is a machine learning strategy that extends Transformation Based Learning by providing automatic template generation. In this work, we propose an ETL approach to the clause identification task. We use the English language corpus of the CoNLL'2001 shared task. The achieved performance is not competitive yet, since the F β=1 of the ETL based system is 80:55, whereas the state-of-the-art system performance is 85:03. Nevertheless, our modeling strategy is very simple, when compared to the state-of-the-art approaches. These first findings indicate that the ETL approach is a promising one for this task. One can enhance its performance by incorporating problem specific knowledge. Additional features can be easily introduced in the ETL model.
KW - Informatics
KW - Entropy guided transformation learning
KW - clause identification
KW - machine learning
KW - CoNLL'2001 corpus
KW - natural language processing
KW - Business informatics
UR - http://www.scopus.com/inward/record.url?scp=77955941946&partnerID=8YFLogxK
U2 - 10.1109/STIL.2009.10
DO - 10.1109/STIL.2009.10
M3 - Article in conference proceedings
AN - SCOPUS:77955941946
SN - 978-1-4244-6008-3
SP - 117
EP - 124
BT - 2009 Seventh Brazilian Symposium in Information and Human Language Technology, STIL 2009
PB - IEEE - Institute of Electrical and Electronics Engineers Inc.
CY - Piscataway
T2 - 7th Brazilian Symposium in Information and Human Language Technology, STIL 2009
Y2 - 8 September 2009 through 11 September 2009
ER -