Implementation details of experiments described in paper
Attention and Lexicon Regularized LSTM for Aspect-based Sentiment Analysis
Experiments implemented with python=3.6, tensorflow=1.5.0, see requirements for more details.
├── configs -> Experiment configurations
├── data -> Datasets
├── ...
├── glove_lookup.parquet -> Glove vectors of corpus vocabulary in compressed format
├── glove_symdict.yml -> Corpus vocabulary
├── *.test.csv -> Training set
├── *.train.csv -> Test set
├── logs
├── models -> Directory for saving trained models
├── src
├── ...
├── models
├── atlstm.py -> AT-LSTM model implementaion (Wang et al.)
├── atlx.py -> ATLX model implementaion
...
Build environment:
pip install -r requirements.txt
Run experiments from project directory (Windows):
AT-LSTM (baseline)
script cross_validate data/processed/SemEval14/SemEval14_train.csv -t data/processed/SemEval14/SemEval14_test.csv -m atlx -e 1 -k 6
ATLX
script cross_validate data/processed/SemEval14/SemEval14_train.csv -t data/processed/SemEval14/SemEval14_test.csv -m atlx -e 3 -k 6
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent