Word Embeddings Benchmarks
Word Embedding Benchmark (web) package is focused on providing methods for easy evaluating and reporting results on common benchmarks (analogy, similarity and categorization).
Research goal of the package is to help drive research in word embeddings by easily accessible reproducible results (as there is a lot of contradictory results in the literature right now). This should also help to answer question if we should devise new methods for evaluating word embeddings.
To evaluate your embedding (converted to word2vec or python dict pickle)
on all fast-running benchmarks execute
See here results for embeddings available in the package.
Warnings and Disclaimers:
- Analogy test does not normalize internally word embeddings.
- Package is currently under development, and we expect within next few months an official release. The main issue that might hit you at the moment is rather long embeddings loading times (especially if you use fetchers).
Please also refer to our recent publication on evaluation methods https://arxiv.org/abs/1702.02170.
- scikit-learn API and conventions
- 18 popular datasets
- 11 word embeddings (word2vec, HPCA, morphoRNNLM, GloVe, LexVec, ConceptNet, HDC/PDC and others)
- methods to solve analogy, similarity and categorization tasks
- Google Analogy
- MSR Analogy
- ESSLI (2b, 2a, 1c)
Note: embeddings are not hosted currently on a proper server, if the download is too slow consider downloading embeddings manually from original sources referred in docstrings.
This package uses setuptools. You can install it running:
python setup.py install
If you have problems during this installation. First you may need to install the dependencies:
pip install -r requirements.txt
If you already have the dependencies listed in
to install in your home directory, use:
python setup.py install --user
To install for all users on Unix/Linux:
python setup.py build sudo python setup.py install
You can also install it in development mode with:
python setup.py develop
See examples folder.
Code is licensed under MIT, however available embeddings distributed within package might be under different license. If you are unsure please reach to authors (references are included in docstrings)