Textual emotion recognition has been a promising research topic in recent years. Many researchers aim to build more accurate and robust emotion detection systems. In this paper, we conduct several experiments to indicate how data pre-processing affects a machine learning method on textual emotion recognition. These experiments are performed on the Vietnamese Social Media Emotion Corpus (UIT-VSMEC) as the benchmark dataset. We explore Vietnamese social media characteristics to propose different pre-processing techniques, and key-clause extraction with emotional context to improve the machine performance on UIT-VSMEC. Our experimental evaluation shows that with appropriate pre-processing techniques based on Vietnamese social media characteristics, Multinomial Logistic Regression (MLR) achieves the best F1-score of 64.40%, a significant improvement of 4.66% over the CNN model built by the authors of UIT-VSMEC (59.74%).
View on arXiv