danieldeutsch / sacrerouge

SacreROUGE is a library dedicated to the use and development of text generation evaluation metrics with an emphasis on summarization.
Apache License 2.0
134 stars 11 forks source link

Ensure ROUGE data is valid #33

Closed danieldeutsch closed 4 years ago

danieldeutsch commented 4 years ago

Successfully setting up ROUGE can sometimes require rebuilding the data files, like so:

https://github.com/danieldeutsch/sacrerouge/blob/c2147cce70d477e175acfbb94063496de3621fe8/.github/workflows/unit-test.yml#L39-L46

If the data files aren't correctly formatted, I believe ROUGE does not crash, but it will be incorrectly calculated. We should (1) detect when this happens to alert the user, and (2) potentially try to fix it ourselves by running the above comands.