Datasets:

wmt

/

wmt18

like 7

Follow

WMT: Workshop on Statistical Machine Translation 16

After investigating this issue, we have discovered that there are issues with the source data file of the Common Crawl corpus: https://www.statmt.org/wmt13/training-parallel-commoncrawl.tgz

• Non-English files contain many English sentences.
• Their "parallel" sentences in English are not aligned: they are uncorrelated with their counterpart

This PR adds a warning to the dataset card explaining this.