InterpretationData
This repository is for our EMNLP 2021 paper: It is Not as Good as You Think! Evaluating Simultaneous Machine Translation on Interpretation Data
There are two stages in our human annotation process: 1) segmentation; 2) ASR error correction. You can find our Interpretation test sets after each stage in noised-interpretation
and clean-interpretation
, respectively. The corresponding Translation test set can be found in clean-translation
.