Classification of Emotions Based on Text and Qualitative Variables

The aim of the paper is to compare the performance of classification of samples based on self-reporting emotions of subjects described by qualitative variables and textual description of situation in which it is experienced. The ISEAR data set was used in the research, which consists of 7666 samples classified in seven classes representing basic emotions (joy, fear, anger, sadness, disgust, shame, and guilt). For the text-based classification logistic regression (LR) was used as well as deep learning methods of convolutional neural networks (CNN), long short-term memory networks (LSTM) and convolutional long short-term memory networks (C-LSTM), while for classification based on qualitative variables LR and multilayer perceptron (MLP) were used. Performance of classification according to the textual description is similar for the LR and deep learning methods ranging from 53.65% in accuracy for CNN to 57.43% for LSTM, while the LR (accuracy of 77.19%) outperforms the MLP (accuracy of 51.83%) in classification according to qualitative variables. Also, we constructed an ensemble of classifiers based on text and qualitative variables showing improvement in performance compared to separate classifiers and accuracy ranging from 63.04% (CNN for text and MLP for qualitative variables) to 80.55% (LR for both text and qualitative variables).