Maple: An emotion-based machine learning chatbot

Maple is a discord.py & Tensorflow-based RNN (Recurrent neural network). She has conversational memory which allows her to continue conversations over multiple messages, and even has sentiment detection which allows her to react emotionally to your messages.

Your initial experience with Maple will vary from person to person as she can generate all different kinds of responses to the same messages. I recommend to keep talking with her on and off (or more if you like) for a couple days and try it out because at times the quality of the responses can be better or worse than the usual. Also, she is better with general conversation and less so with specific questions. She is not capable of long-term memory and will not always provide good responses to specific/objective questions. For example, you might ask a subjective question like "How was your day today?" or "Did you have fun at the party?" and she will usually do very well. But a question like "What is 2+2?" or "What is my name?" will not work nearly as well.

Surprisingly, Maple is quite capable with roleplaying, though this was not intentional. The RNN happened to end up being good at it even though I did not intend to give her that capability.

She will respond to messages on servers by mentioning her, and works just fine without mentions via DMs. Note that I did not add traditional commands because Maple is meant to work with any text input (as long as she is mentioned.) Commands would not work well given Maple's chatbot nature.

Maple is multilingual; just speak in your native language and she will respond in kind. Note that results will not always be amazing and there might be grammatical errors here and there.

Multilingualism has been restored! I'm happy to have it working again.

The profile picture is AI-generated too! The person pictured does not0exist. They were created by the people at https://generated.photos/

Check out Maple's Version 3codebase on Github if you wish to run her yourself. Note that the setup process is advanced and has many dependencies, most of which are easiest to run on Ubuntu or the Linux distribution of your choice. She is based off of lukalabs' Cakechat program.