Boosting-based ensemble learning with penalty setting profiles for automatic Thai unknown word recognition

Authors:
Jakkrit TeCho;Cholwich Nattee;Thanaruk Theeramunkong
Affiliations:
School of Information, Computer and Communication Technology, Sirindhorn International Institute of Technology, Thammasat University, Thailand;School of Information, Computer and Communication Technology, Sirindhorn International Institute of Technology, Thammasat University, Thailand;School of Information, Computer and Communication Technology, Sirindhorn International Institute of Technology, Thammasat University, Thailand
Venue:
ICCCI'10 Proceedings of the Second international conference on Computational collective intelligence: technologies and applications - Volume Part II
Year:
2010

Citing 3
Cited 0

Ensemble Methods in Machine Learning

MCS '00 Proceedings of the First International Workshop on Multiple Classifier Systems
The automatic extraction of open compounds from text corpora

COLING '96 Proceedings of the 16th conference on Computational linguistics - Volume 2
A collaborative framework for collecting Thai unknown words from the web

COLING-ACL '06 Proceedings of the COLING/ACL on Main conference poster sessions

Quantified Score

Hi-index	0.00

Visualization

Abstract

A boosting-based ensemble learning can be used to improve classification accuracy by using multiple classification models constructing to cope with errors obtained from preceding steps. This paper presents an application of the boosting-based ensemble learning with penalty setting profiles on automatic unknown word recognition in Thai. Treating a sequential task as a non-sequential problem requires us to rank a set of generated candidates for a potential unknown word position. Since the correct candidate might not located at the highest rank among those candidates in the set, the proposed method provides penalties, in the form of a penalty setting profile, to improper ranking in order to reconstruct the succeeding classification model. In addition a number of alternative penalty setting profiles are introduced and their performances are compared on the task of extracting unknown words from a large Thai medical text. Using the naïve Bayes as the base classifier for ensemble learning, the proposed method achieves the accuracy of 89.24%, which is an improvement of 9.91%, 7.54%, 5.25% over conventional naïve Bayes, nonensemble version, and flat penalty setting profile.