Emotion Classification Using Support Vector Machine

Abstract

Emotions have always been significant in how people connect with one another in daily life. This feeling has recently gained importance in human-computer interactions as well. Emotions may be expressed by humans through writing, voice, and facial expressions. Text-to-emotion recognition is a classification task with predefined emotion labels. This study uses the following seven categories of emotions: joy, anger, sadness, fear, disgust, shame and guilt. These categories are drawn from the ISEAR (International Survey on Emotion Antecedent and Reaction) dataset, which consists of 7666 original lines of English sentences with emotion labels. To determine which of the three support vector machine technique kernels—linear, RBF, and polynomial—performed best for text categorization, a comparison of the kernels was also conducted. Rotating models that are created from the outcomes of training on training data employ a variety of metrics. 61.3% of the linear kernel with parameter C = 0.5, 60.3% of the RBF kernel with parameter C = 1 and γ = 2, and 57.7% of the polynomial kernel with parameter C = 5, γ = 0.8, and degree = 2 were the accuracy values obtained for each kernel based on the test results. It has been demonstrated that the linear kernel performs better in text categorization than other kernels.