toxic-message-classification

This app receives incoming messages from a Slack channel, stores this real-time chat data in a Deephaven streaming table, predicts a probability of toxicity for each message, and sends a notification if a message is classified as toxic.

Components

General

• start.sh - A helper script to launch the application.
• docker-compose.yml - The Docker compose file that defines the Deephaven images.
• Dockerfile - The Dockerfile for the application. Simply extends Deephaven's base image with dependencies and app-mode scripts.
• requirements.txt - The Python dependencies for the app.

Files

• main.py - main Python script
• app.app - The Deephaven App Mode config file

High level overview

This app checks if a new message posted to a Slack channel reads as toxic. If so, a bot sends a warning message to the channel.

The process contains 3 steps:

• receive slack events using the Slack Real Time Messaging API and store real-time chat data in a Deephaven streaming table
• predict a probability of toxicity for each message
• send a notification if a message is classified as toxic

How to run

Run sh start.sh and open http://localhost:10000/ide/

Note

This repository was built for Deephaven Community Core v0.15.1. No guarantee of forwards or backwards compatibility is given.