[Unsupervised and Very Low-res] Preliminary results

7 views
Skip to first unread message

Jindřich Libovický

unread,
Jul 26, 2021, 3:31:56 PM7/26/21
to Workshop on Statistical Machine Translation

Hi folks,

The preliminary results of the task are here:

We scored the systems using BLEU, chrF and BERTScore (using XLM-R for Russian and German and mBERT for Chuvash) and for each system computed how many systems were significantly outperformed in given the metrics using bootstrap resampling. (Please note we might change/improve/fix the significance testing before the final findings and include some other metrics.)

Best,
Jindřich


Best,
Jindřich
Reply all
Reply to author
Forward
0 new messages