Evaluating the Impact of Dataset Variability on Malayalam Emotion Detection System

Text emotion detection has become an increasingly important area of research, with applications ranging from sentiment analysis to enhanced human-computer interaction. The performance of emotion detection systems can, however, vary widely depending on the datasets used for training and evaluation. An emotion detection (ED) system is evaluated, comparing its performance across two datasets that differ in origin, structure, and emotional category, is presented in this paper. We employed a range of evaluation metrics, including accuracy, precision, recall, and F1-score, to analyze the system's performance. The results highlight the impact of dataset char-acteristics on the effectiveness of ED models, revealing notable differences between the two datasets. The performance of the ED system is significantly influenced by factors such as the diversity of emotions and the quality of annotations. This insight contributes to the development of more adaptable and reliable ED systems across a variety of domains. A discussion of future research directions is included at the end of the paper.