This dir displays the human evaluation file for self-estimation.
There are five dirs respectively for zh-en, fr-en, ja-en, en-de and zh-en multi-domain tests.
Each has six files:
xx.src.txt: the source sentences.
xx.ref.txt: the human references given by the testset.
xx.trans.txt: the machine translation generated by CANMT.
xx.human_scores.txt: the translation quality score annotated by human translators.
xx.averaged_human_scores.txt: the averaged human quality scores.
xx.canmt_scores.txt: the predicted score by CANMT.
