-
DL approach using Bi-directional LSTM: sentiment-analysis.py
-
ML approach using various effective classification models: sentiment-analysis-ml.py
-
extension.csv : additional data available for usage on top of train.csv
Note:
- Dataset is slightly skewed to classes 4/5, Matthew's correlation coefficient might be a better metric over 'accuracy'
- Reviews contains malaysian and indonesian melayu words