Cognate Transformer for Automatic Phonological Reconstruction

Usage:

python run.py <method> Where <method> can be reflex, proto-cog or proto-nmt

Results are generated in results/

Pip dependencies are listed in requirements.txt

Additionally, the following line is to be added to trainer_seq2seq.py file from transformers library for NMT to work correctly.

if "langs" in inputs: gen_kwargs["langs"] = inputs.get("langs", None)

The line number in the file would be at 189 (in transformers v4.24.0) and the file location would be like .../anaconda3/lib/site-packages/transformers or depends on the environment.

Citation

V.S.D.S.Mahesh Akavarapu and Arnab Bhattacharya. 2023. Cognate Transformer for Automated Phonological Reconstruction and Cognate Reflex Prediction. In Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing, pages 6852-6862, Singapore. Association for Computational Linguistics.link