Public Evaluation Data

[bminixhofer]

Is the evaluation data available / can you make it available? Sorry if it already is and I missed it.

Specifically referring to this part in the paper:

We construct test sets from the WMT News Translation test sets (Barrault et al., 2020), which provides for decent-size test sets in many languages. We manually corrected all sentence segmentations. While some sets were already well segmented, some more recent years were extremely under-segmented.

Having the evaluation data as used in your paper would make it much easier to build on your work. Thanks!

[bminixhofer]

Ok, nevermind, I found it :) For future reference, the data is available here: https://github.com/rewicks/ersatz-test-suite.