Can you give more details of reproduce the paper results?

facebookresearch / mlqe

We release a dataset based on Wikipedia sentences and the corresponding translations in 6 different languages along with the scores (scale 1 to 100) generated though human evaluations that represent the quality of the translations.Paper Title Unsupervised Quality Estimation for Neural Machine Translation

Creative Commons Attribution Share Alike 4.0 International

80 stars 14 forks source link

Can you give more details of reproduce the paper results? #3

Closed maohbao closed 3 years ago

maohbao commented 3 years ago

Hi Marina,

Thank you very much for the job. I want to reproduce the results, but can you give me more details about how to do this?

After I installed fairseq, how to get the QE score for an input sentence during translation? I shall modify the transformer decoder code of fairseq to get more information during decode, is not it? Can you provide more information about how to do this? Thank you very much!

Bao

mfomicheva commented 3 years ago

Hi Bao, Thanks a lot for your interest. You have instructions and some scripts to reproduce the results here: https://github.com/pytorch/fairseq/tree/master/examples/unsupervised_quality_estimation Please, let me know if you have any further questions

maohbao commented 3 years ago

Thank you very much, I will try it!