Sentiment Analysis by radical(of Japanese kanji character)-level encoder.
This encoder refers to following paper.
- OS: Ubuntu 16.04.4 LTS
- GPU: GeForce GTX 1080
- Python: 3.6.5
- Chainer: 4.2.0
- MeCab: 0.996 (split sentence to words)
Directory configuration as follows.
├── create_vocab.py
├── radical_classifier.py
├── char2radical.py
└── datasets
├── radical_dic.csv
└── set [train/test datasets]
Make vocabulary-dictionary of your datasets with the ./datasets/radical_dic.csv
python create_vocab.py
Edit Hyper-parameter, Run training script with the generated file.
python radical_classifier.py