Code accompanying the EMNLP 2023 publication "Recurrent Neural Language Models as Probabilistic Finite-state Automata".

Getting started with the code

Clone the repository:

$ git clone [email protected]:rycolab/weighted-minsky.git
$ cd weighted-minsky

At this point it may be beneficial to create a new Python virtual environment. There are multiple solutions for this step, including Miniconda. We aim at Python 3.10 version and above.

Then you install the package in editable mode:

$ pip install -e .

We use black and flake8 to lint the code, pytype to check whether the types agree, and pytest to unit test the code.

---

Citation

@article{svete2023recurrent,
      title={Recurrent Neural Language Models as Probabilistic Finite-state Automata}, 
      author={Anej Svete and Ryan Cotterell},
      year={2023},
      eprint={2310.05161},
      archivePrefix={arXiv},
      primaryClass={cs.CL},
      journal = {arXiv preprint arXiv:2310.05161},
      url={https://arxiv.org/abs/2310.05161}
}