naver-ai/MetricMT
The official code repository for MetricMT - a reward optimization method for NMT with learned metrics
When evaluating machine translation systems, the common BLEU metric often fails to align with human judgment. This project provides a way for machine translation developers to fine-tune their translation models using human-aligned metrics like BLEURT. The input is an existing Neural Machine Translation (NMT) model, and the output is an NMT model optimized to produce translations that humans perceive as higher quality. This is for machine translation researchers and engineers who want to improve the perceived quality of their translation systems.
No commits in the last 6 months.
Use this if you are a machine translation developer seeking to improve your NMT model's output quality according to human perception rather than relying solely on traditional automated metrics.
Not ideal if you are a casual user looking for an off-the-shelf translation tool, or if you are not working with Neural Machine Translation models.
Stars
25
Forks
2
Language
—
License
MIT
Category
Last pushed
Apr 24, 2021
Commits (30d)
0
Get this data via API
curl "https://pt-edge.onrender.com/api/v1/quality/ml-frameworks/naver-ai/MetricMT"
Open to everyone — 100 requests/day, no key needed. Get a free key for 1,000/day.
Higher-rated alternatives
facebookresearch/fairseq2
FAIR Sequence Modeling Toolkit 2
lhotse-speech/lhotse
Tools for handling multimodal data in machine learning projects.
google/sequence-layers
A neural network layer API and library for sequence modeling, designed for easy creation of...
awslabs/sockeye
Sequence-to-sequence framework with a focus on Neural Machine Translation based on PyTorch
OpenNMT/OpenNMT-tf
Neural machine translation and sequence learning using TensorFlow