Disambiguated skip-gram model

This is a TensorFlow 1.x implementation of the disambiguated skip-gram model introduced in [1].

Instalation

The dependencies are listed in requirements.txt. We recommend creating a Python 2.x virtual environment and then installing the required packages using:

pip install -r requirements.txt

Training

To reproduce the results from [1] you first need to download the Wesbury Lab Wikipedia corpus and unpack it to the datasets directory. You can then run the script that will prepare the training data, train the model and evaluate it against WSI benchmarks:

./prepare_train_and_test.sh

On a 24-core machine with 128 GB of RAM this script takes around 2.5 days to train the model.

Disclaimer

This code is intended for replication of the published results and should not be used for commercial purposes.

References

[1] Karol Grzegorczyk, Marcin Kurdziel, Disambiguated skip-gram model, EMNLP 2018, appendix