Emotion Detection in Persian Text; A Machine Learning Model

This study aimed to develop a computational model for recognition of emotion in Persian text as a supervised machine learning problem. We considered Pluthchik emotion model as supervised learning criteria and Support Vector Machine (SVM) as baseline classifier. We also used NRC lexicon and contextual features as training data and components of the model. One hundred selected texts including political-social newspaper editorials were used to test the model (real terms). Also in this study, the "support vector machine" algorithm was used as the learning classifier and four indicators of accuracy, accuracy, f-score and recall were used to evaluate the model. The results show that the efficiency of the model in detecting different emotions varies from 79% to 98% and mean presision of the model for all classes was 84%. Using all indexes, the classifier showed more performance in joy category than other 7 types. The results of this study show that using emotion-based approach, supervised learning and minimal contextual features can be useful in automatic identification of emotions. It also showed that a combination of lexical resource and contextual features can be used as learning base for a SVM model.