Home
After cloning the repository, you should create a directory logs to store log files.
The version of learning code that we reported in the paper is main.py. You could try other versions if you want.
You would need to install numpy and tensorflow. Tensorflow version 0.11 doesn't support certain calculation I used in the code in GPU, so you would need to run the code on CPU.
Description of data.
https://github.com/tuandnvn/ecat_learning/wiki/Data
Description of algorithm.
https://github.com/tuandnvn/ecat_learning/wiki/Algorithm
Here is how you can replicate the experiments.
python main.py
Script's usage:
usage: [-h] [-t] [-m MODEL]
A script to train and test using LSTM-CRF
optional arguments:
-h, --help show this help message and exit
-t, --test If set, it is TEST, by default it is TRAIN
-m MODEL, --model MODEL
Where to save the model or to load the model. By
default, it is saved to the log dir
This would produce a directory named logs/run_year_month_date_hour_minute_second with the following structure:
logs/run_[year]_[month]_[date]_[hour]_[minute]_[second]
├── checkpoint
├── main.py
├── logs.log
├── model.ckpt
└── model.ckpt.meta
After that you can reuse the trained model at model.ckpt to rerun the test:
python main.py -t -m PATH_TO_MODEL_FILE
The code and the data is released under the GNU General Public License