Hi Sandeep
Bleu-A is scored with the A reference, bleu-B with the B
reference and bleu-all with both references - if there are two
references. All systems are scored with sacrebleu 13a
tokenisation, except for Japanese (char-based) and Chinese (zh
tokenisation).
I should emphasise that the automatic scores are just for
guidance. We will publish only human evaluation in the overview,
best
Barry
This email was sent to you by someone outside the University.You should only click on links or attachments if you are certain that the email is genuine and the content is safe.
--
You received this message because you are subscribed to the Google Groups "Workshop on Statistical Machine Translation" group.
To unsubscribe from this group and stop receiving emails from it, send an email to wmt-tasks+...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/wmt-tasks/366e762c-1235-47a7-8c9e-97e79ae72385n%40googlegroups.com.
Hi Jeremy
This should only happen for lines in the testsuites documents, which sometimes do not have a reference. You can add the --no-testsuites option to wmt-unwrap in order to ignore the test suites,
best
Barry
To view this discussion on the web visit https://groups.google.com/d/msgid/wmt-tasks/0fa4066d-44a9-4a32-8092-72fb3876721an%40googlegroups.com.