This repo is the implementation of Transformation Networks for Target-Oriented Sentiment Classification (Xin Li, Lidong Bing, Wai Lam, Bei Shi, 2018) which is accepted by ACL 2018.
Before training the model, you need to get the embedding and dataset.
- Glove embedding: https://nlp.stanford.edu/projects/glove/ (glove.840B.300d)
- Dataset: https://github.com/lixin4ever/TNet
or you can use prepare_data.sh to download the laptop dataset and the embedding automately.
. ./prepare_data.sh
- add the
prepare_data.shscript. - fix the performance problem (forget shuffle on each epoch.)
- Python 3.6.x
- tensorflow 1.12.0
- numpy 1.15.3
- tqdm 4.23.4
pip install -r requirements.txt
python train.py
tensorboard --logdir=log --host=<host> --port=<port>