Removing samples containing profanity

[vmurahari3]

Do you think it makes sense to remove samples containing profanity?

[matthen]

In general there is a lot of questionable language in the reddit dataset, as it is totally unfiltered and we are including all subreddits including 'nsfw' ones. It is still natural language, and a potentially useful learning signal, though of course we need to be careful how the resulting model is used.

We could perhaps add flags to the pipeline for filtering based on the nsfw label etc. These would be off by default.