Learning Neural Network Ensemble for Practical Text Classification

Abstract

Automated text classification has been considered as an important method to manage and process a huge amount of documents in digital forms that are widespread and continuously increasing. Recently, text classification has been applied by machine learning technologies such as k-nearest neighbor, decision tree, support vector machine, and neural networks. However, most of the investigations in text classification are studied not on real data but on well-organized text corpus, and do not show their usefulness. This paper suggests and analyzes text classification method for a real application, FAQ text classification task, by combining multiple classifiers. We propose two methods of combining multiple neural networks that improve performance by maximum combining and neural network combining. Experimental results show the usefulness of proposed methods for real application domain.