DiscordWordCloud

⚠️This version is outdated, I've rewrote the bot in Rust here https://github.com/Inspirateur/wordy ⚠️
A Python 3.9 discord bot that generates a word cloud (now with emojis !) for each discord user.

How to run it

Register Discord Application

• go to https://discordapp.com/developers/applications/ create your app
  • add a User Bot to it and paste its Token in token.txt
  • enable SERVER MEMBERS INTENT and MESSAGE CONTENT INTENT in the bot tab
  • invite the bot with https://discord.com/api/oauth2/authorize?client_id=CLIENT_ID&permissions=0&scope=bot%20applications.commands replace CLIENT_ID with the Client ID of your app

Run Locally

• Clone the project wherever you want
• Install Python 3.9 from https://www.python.org/ (you can probably use an older version too)
• Install the requirements listed in requirements.txt with pip
  • use python -m pip install -r requirements.txt at the root of the project
  • consult this guide if you need help with pip https://packaging.python.org/tutorials/installing-packages/
• add a token.txt file at the root of the project
• run main.py with Python 3.9

Run with Docker

docker run -e TOKEN=TOKEN_FROM_ABOVE_GOES_HERE mrhazel/discord-word-cloud

If you got any questions about this project, feel free to DM Inspi#8989 on Discord.

Features

• ;load (days) use this to load messages up to x days on your server
• ;cloud to generate your word cloud, you can also tag someone in the command to generate their word cloud
• ;emojis to get the custom emojis usage of the server

Pictures

How it works

Here's a step by step of how the bot makes a wordcloud:

• after ;load
  • for each word w and user u, compute p(w|u), the probability of u writing w (this data is separated between servers)
• after ;cloud
  • for each word w and the user u for which the cloud is, compute p(w|u)/p(w), this quantifies how much u favors w compared to everyone else
  • uses the WordCloud lib to make a word cloud image, scaling the words in proportion to their scores (they are placed randomly)

Some notes regarding this model:

• This model as-is would be filled with typos, bits of url and words that have only been written by u because p(w|u)/p(w) would be high; to counter this, a value α is added to every p(w|u) as if each user had at least an α probability of writing any word. This regularisation has the downside of putting some words that were never written by u in the word cloud.
• A strong point of this model is that it works in any language and stop-words such as and, the, a, etc. are often excluded because they are not favored by anyone in particular.
• One limitation of the model is that it does not handle expressions bigger than a single word or emoji.

The code is written so that implementing a new model is easy, by subclassing WCModel in wcmodel.py, and importing your custom class instead of WCBaseline in main.py !

Powered by:

• Pycord: https://github.com/Pycord-Development/pycord
• WordCloud: https://github.com/amueller/word_cloud
• Twemoji: https://twemoji.twitter.com/