In this project, we are trying to explore the effectiveness of machine learning models for multiclass emotion classification on a textual dataset. The dataset comes from Kaggle (https://www.kaggle.com/datasets/praveengovi/emotions-dataset-for-nlp), which contains labeled text samples belonging to six distinct emotions (sadness, fear, angry, joy, love, surprise), and the goal is to develop a model that accurately classifies these emotions.

• LSTM+CNN model

  This model is implemented in the Code/emotions_LSTM.py. Go into the folder and run the code directly.

• Transform (DistilBert) model

  This model is implemented in the Code/transformer_model.py. Go into the code and make sure the parameter of model_bert_lstm = False in line 32.

• Transform model with LSTM head

  This model is implemented in the Code/transformer_model.py. Go into the code and make sure the parameter of model_bert_lstm = True in line 32.

The model interpretation only works in Transform (DistilBert) model. So make sure model_bert_lstm = False if you want to get the results of model interpretation.

• Sample LIME result

• Sample SHAP result