Discriminative training and support vector machine for natural language call routing

Abstract

In natural language call routing, callers are routed to desired departments based on natural spoken responses to an open-ended "How may I direct your call?" prompt. Natural language call classification can be performed using support vector machines (SVMs) or the popular vector-based model used in information retrieval. We recently demonstrate how discriminative training is powerful to improve any parameterized vector-based classifier to achieve minimum classification error. Discriminative training minimizes the classification error by increasing the score separation of the correct from competing documents. It makes the classifier robust to feature selection, enabling fully automated training without the injection of human expert knowledge. Support vector machines received also a lot of attention in the machine learning community. They have often achieved better performance than customized neuronal network and state-of-the-art baseline classifiers. We investigate in this paper the classification power of SVMs and discriminative training approaches on natural language call routing. Experiments are reported for a banking call routing and for Switchboard topic identification task. Results show that the application of discriminative training on vector-based model outperforms SVMs by 7% on spoken data.