In this paper, the performance of Support Vector Machine (SVM) and Decision Tree (DT) in classifying emotions from Malay folklores is presented. This work is the continuation of our storytelling speech synthesis work to add emotions for a more natural storytelling. A total of 100 documents from children short stories are collected and used as the datasets of the text-based emotion recognition experiment. Term Frequency-Inverse Document Frequency (TF-IDF) is extracted from the text documents and classified using SVM and DT. Four types of common emotions, which are happy, angry, fearful and sad are classified using the two classifiers. Results showed that DT outperformed SVM by more than 22.2% accuracy rate. However, the overall emotion recognition is only at moderate rate suggesting an improvement is needed in future work. The accuracy of the emotion recognition should be improved in future studies by using semantic feature extractors or by incorporating deep learning for classification.

No abstract