Deep Learning for Second Language Acquisition Modeling

Using RNNs for Second Language Acquisition Modeling based on the 2018 competition by Duolingo http://sharedtask.duolingo.com/

Requirements

Python 3.5/7 and other libraries. To install everything needed please run

pip install requirements.txt

Execute

1. Unpack data

In the data folder give permissions to the unpacking script by running:

chmod 777 unpack_data.sh

then execute it to unpack the data in put it in a folder.

./unpack_data.sh

In the starter_code folder, unpack the glove.6B.50d.txt.tar.gz that contains pretrained word embeddings:

tar -xvzf glove.6B.50d.txt.tar.gz

2. Run count_features.py to go over all data and make a dictionary of the features and how many times each value appears (important for feature selection)

python3 count_features.py

3. Execute either lstm.py for a default run of a model or experiments.py to run different comparative experiments.

python3 lstm.py

or run baseline.py for Duolingo's baseline Logistic Regression model.