Detect intensity for 4 different emotions from the tweet content.
Key techniques used for feature engineering:
- Emoji extraction and interpretation
- Hashtag extraction and segmentation
- Spelling correction
- Accent/Diacritic removal
- Emotion synonym similarity/antonym dissimilarity using word embeddings
- Cuss word detection
- Extracting other token properties
notebooks/pipeline.ipynb: Constructing the tokenizer and feature engineering.
notebooks/training.ipynb: Training and evaluation of all classifiers.