Potentially as another provider for the sentiment analysis rule or as a separate rule entirely.
Would provide another signal for comment toxicity. Things to think about WRT where to add...
MHS results only provide a confidence score (0 to 1) which doesn't mesh well with sentiment score (-1 to 1) mostly because sentiment results tell positive/negative and their score is just "how confident is this toxic" -- seems to point towards an independent rule.
On the other hand, requiring a separate rule to be configured/added introduces more complexity. But it also allows for more granularity...
https://moderatehatespeech.com/ https://www.reddit.com/r/redditdev/comments/xdscbo/updated_bot_backed_by_moderationoriented_ml_for/ https://moderatehatespeech.com/docs/
Potentially as another provider for the sentiment analysis rule or as a separate rule entirely.
Would provide another signal for comment toxicity. Things to think about WRT where to add...
MHS results only provide a confidence score (0 to 1) which doesn't mesh well with sentiment score (-1 to 1) mostly because sentiment results tell positive/negative and their score is just "how confident is this toxic" -- seems to point towards an independent rule.
On the other hand, requiring a separate rule to be configured/added introduces more complexity. But it also allows for more granularity...