Topic identification for Turkish call center records

In this paper, feature extraction and classification of text contexts which are obtained from call center audio records by means of speech recognition is investigated by using various pattern recognition methods. Vector space model, latent semantic analysis and iterative residual rescaling methods are used for extracting text features while K-means classifier, support vector machines and artificial neural networks are preferred for classifying texts according to previously defined topics. Robustness of these pattern recognition methods is tested by investigating the effect of changes in voice recognition performance on classification of textual content.