The problem is treated in two phases :
- Aspect extraction consists of extracting and classifying aspect targets in the review
- Sentiment analysis consists of assigning a tonality (negative, neutral, positive) to each extracted aspect target
pip install -r requirements.txt
python api.py
curl --request GET --url http://localhost:9090/api-sentiment-1/doc/
- Build vocab from the data according to the config in
sequence_tagging/model/config.py
.
python sequence_tagging/build_data.py
- Train the model with
python sequence_tagging/train.py
- Evaluate and interact with the model with
python sequence_tagging/evaluate.py
Train/test a model according to the config in mem_absa/config_mem.py
python mem_absa/train_test.py --show True