Moderates discord messages and images.
More and more people are using discord form everyday tasks, classroom-activity to office meetings. And places lie these should be safe for everyone. With the incresing number of people coming to discord for education/tech/social communities, it's hard to keep the server safe for everybody. Sometimes some people send inappropriate messages which might hurt the sentiment of others.
Thats why we created makeitsafe.space discord bot.
What it does
- Text Moderation
If User send some inappropriate message then that message will deleted after 5 seconds atomatically. Currently this bot detect spam, insult, profanity, flirtation, threat, sexually inappropriate message.
- Image Moderation
If user upload indecorous image then image moderation library classsify image in drawing, hentai, neutral, porn, sexy categories. Based on that inappropriate image will deleted automatically.
How we built it
We used the discord.js api for interacting with discord. We tried perspective API, google cloud ML apis, Azure content moderation APIs and NSFW.js. We finalized on perspective API for text moderation and nsfw.js for media moderation. We developed stats server using node/express.
Perspective was created by Jigsaw and Google’s Counter Abuse Technology team in a collaborative research project called Conversation-AI.
For Image Moderation we use NSFW.js Which under the hood uses tensorflow.js
Challenges we ran into
- figuring out best false-positive rate by changing prediction probablity.
Accomplishments that we're proud of
- working bot server.
- perspective api configuration
What we learned
- making discord api
- adding Perspective API
- NSFW library
What's next for test
- get individual person stats