Decision tree decomposition-based complex feature selection for text chunking

Incorporating a method of feature selection into a classification model often provides a number of advantages. In this paper we propose a new feature selection method based on the discriminative perspective of improving the classification accuracy. The feature selection method is developed for a classification model for text chunking. For effective feature selection, we utilize a decision tree as an intermediate feature space inducer. To select a more compact feature set with less computational load, we organized a partially ordered feature space according to the IGR distribution of features. Experimental results show that: (1) the computational complexity on high-dimensional feature space can be reduced by selecting features based on the decision tree decomposition; (2) the text chunking system using the proposed feature selection can significantly improve the performance compared with a decision tree classifier.