BiVERT: Bidirectional Vocabulary Evaluation using Relations for Machine Translation

BIVERT arxiv / BabelNet

Introduction

BiVERT is a new evaluation method for machine translation. It uses BabelNet. This repository contains the code for BiVERT.

How To Use

This code uses the full extended version of BabelNet. Please refer to BabelNet to get full access.

See babelnet-server -> Main.java for localhost API server to access babelnet files.

Main.py -> Function for creating BiVert category data and training.

For data files please refer to WMT MQM Metrics Task.

Key Concepts

Citations

If you find our code/paper/method helpful, please cite our paper 📝 and star us ⭐️！

@misc{cherf2024bivertbidirectionalvocabularyevaluation,
      title={BiVert: Bidirectional Vocabulary Evaluation using Relations for Machine Translation}, 
      author={Carinne Cherf and Yuval Pinter},
      year={2024},
      eprint={2403.03521},
      archivePrefix={arXiv},
      primaryClass={cs.CL},
      url={https://arxiv.org/abs/2403.03521}, 
}