tomfisher/distance-embed

Code the paper titled Unsupervised Sentence-embeddings by Manifold Approximation and Projection published at EACL 2021 (https://arxiv.org/abs/2102.03795) by Subhradeep Kayal.

Citation

If you find this paper useful for your research, please consider citing the paper:

@inproceedings{kayal-2021-unsupervised,
    title = "Unsupervised Sentence-embeddings by Manifold Approximation and Projection",
    author = "Kayal, Subhradeep",
    booktitle = "Proceedings of the 16th Conference of the European Chapter of the Association for Computational Linguistics: Main Volume",
    month = apr,
    year = "2021",
    address = "Online",
    publisher = "Association for Computational Linguistics",
    url = "https://www.aclweb.org/anthology/2021.eacl-main.1",
    pages = "1--11"
}

Installing dependencies

• all the dependencies have been explicitly mentioned in requirements.txt
• to prepare the environment, run prepare-server.sh
• this will:
  • install python2 and 3
  • install swig
  • create and activate a virtual environment named distemb
  • install the packages in requirements.txt
  • install this particular package
  • create necessary folders
  • fetch the GoogleNews-vectors-negative300.bin.gz corpus
  • make the emd package

To run all experiments

• go to /lib/shell
• run, in order:
  • calc_wmd_dist.sh
  • calc_all_dist.sh
  • calc_all_embeddings.sh
  • calc_all_dct.sh, calc_all_eigensent.sh, calc_all_powermeans.sh, calc_all_sbert.sh can be run in any order
  • evaluate_all.sh