Low Accuracy for Emotion Classification

I have been working on Emotion Classification for more than 1 year but I have been getting 50- 57% accuracy for Tree Ensemble ,Gradient Boost and XG Boost .I am unable to determine reason for low accuracy ,other work for same emotion model have 80% accuracy.
Emotion_Classification.knwf (1.1 MB)

mood-mer2000+lyrics.xlsx (1.3 MB)

Have you tried different models? What do other models do different? Have you already checked? Maybe word embedding is better then TFIDF. Maybe Hyperparameter tuning gets you better results.
br

I haven’t tried Hyperparameter tuning and word embedding yet, I will try it