Skip to content

Getting interpretable dimensions in word embedding spaces.

License

Notifications You must be signed in to change notification settings

pdufter/densray

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

5 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Introduction

This repository compares different methods of obtaining interpretable dimension in word embedding spaces.

More specifically it compares:

  • Densifier
  • DensRay: A method closely related to Densifier, but computable in closed form.
  • Support Vector Machines / Regression
  • Linear / Logistic Regression.

The evaluation tasks are lexicon induction and set-based word analogy.

For more details see the Paper.

Note that this repo does not include an implementation of the Densifier, but relies on the original Matlab implementation by the authors of Densifier.

Usage

For an example how to use the code see example.sh.

References

If you use the code, please cite

@article{dufter2019analytical,
  title={Analytical Methods for Interpretable Ultradense Word Embeddings},
  author={Dufter, Philipp and Sch{\"u}tze, Hinrich},
  journal={arXiv preprint arXiv:1904.08654},
  year={2019}
}

About

Getting interpretable dimensions in word embedding spaces.

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published