Stacked Generalization for Topic Classification of Spoken Inquiries

概要

論文の詳細を見る
Stacked generalization is a method that allows to combine outputs of multiple classifiers using a second-level classification, minimizing the generalization error of first-level classifiers and achieving greater predictive accuracy. In a previous work, we compared the performance of different methods in the classification in topics of spoken inquiries in Japanese received by a guidance system, where support vector machines and maximum entropy presented the best performances. In the present work, we employ a stacked generalization scheme that uses predictions of support vector machines and maximum entropy classifiers as input for a second-level classification using linear support vector machines. Experimental results show an improvement in the performance from 94.2% to 94.8% in the classification of automatic speech recognition results from adults and from 88.3% to 89.1% for children, when using stacked generalization in comparison to the individual performance of the methods.
2011-01-28