naacl16-demo

Data, classifiers, and notebooks for the LIME demonstration at NAACL 2016

Clone the repo and copy data

git clone git@github.com:UW-MODE/naacl16-demo.git
cd naacl16-demo

Download the models and copy models.tar.gz to naacl-demo

###Optional: start a virtualenv

virtualenv ENV
source ENV/bin/activate

Install and unpack stuff:

pip install lasagne lime flask theano 
tar xvzf models.tar.gz

Run the first demo (explanation, compare models):

python server.py

Run the second demo (feature engineering)

cd cleaning
python server.py

How we trained the models / data / etc

getting embeddings:ver Getting word2vec stuff notebook
Download the data from here
Unpack the data: tar xvzf datasets.tar.gz

Run the following

  python learn_models.py                                                                    
  python run_submodular.py
  cd cleaning
  python generage_json.py