A tool for matching word sense embeddings with synsets of lexical resources: the method helps to make word sense embeddings interpretable.

See the website for description of the method: http://tudarmstadt-lt.github.io/vec2synset

1. Clone repository:

git clone https://github.com/tudarmstadt-lt/vec2synset.git

2. Install dependencies:

pip install -r requirements.txt

3. Download resources into the repository (100 Mb):

cd vec2synset/data && wget http://panchenko.me/data/joint/adagram/babelnet-bow-5190.pkl

4. Run a test matching (50 words):

python run.py

You should get the file with matched AdaGram and BabelNet senses similar to the one in data/voc-50.csv.match_example.csv.