search

behaviorbot / sentiment-bot

Replies to toxic comments with a maintainer designated reply and a link to the repo's code of conduct
https://github.com/apps/sentiment-bot
MIT License
83 stars 12 forks source link

question(docs): toxicity threshold #24

Closed Kristinita closed 6 years ago

Kristinita commented 6 years ago

1. Summary

It would be nice, if sentimentBotToxicityThreshold key description will be more understandable.

2. Details

Sentiment Bot description:

*Required* toxicity threshold between 0 and .99 with the higher numbers being the most toxic
Anything higher than this threshold will be marked as toxic and commented on

What is “being the most toxic”?

Did I understand correctly that:

?

Or vice versa?

Thanks.

hiimbex commented 6 years ago

The perspective api works by giving any string a toxicity rating. The higher the number, the more toxic the comment was. Therefore if you want the bot to comment only on comments that have a toxicity score from the api of .7 or higher, you'd set that value to .7. If you set this score to .0001 or something very low, the bot might comment on everything, even normal statements!

More details on how the perspective api works here: https://github.com/conversationai/perspectiveapi/blob/master/api_reference.md