Interpreting Embedding Spaces by Conceptualization

Steps to run this code:

1. create an environment with environment.yml in the additional files
2. cd src
3. CUDA_VISIBLE_DEVICES="" python run_tests.py -p ../additional_files/ --human_and_model_evaluation --classification_test --triplets_test --example_creation --model_application --full_llm_explained

To see help about the arguments run python run_tests.py -h

You can also generate to a sentence it's top CES concepts using InterpretingEmbeddingSpacesByConceptualization.ipynb notebook

How to cite this work:

This work has been accepted to the 2023 EMNLP conference. If you use this code, please cite our paper:

  title={Interpreting Embedding Spaces by Conceptualization},
  author={Simhi, Adi and Markovitch, Shaul},
  journal={arXiv preprint arXiv:2209.00445},
  year={2022}
}