Random sentence generator trained on KonoSuba light novel using Markov chain.
Also included is a FastAPI webserver using Aqua to generate random sentences.
Training data is based on the one used by Kazuma.
Normal sentences:
- If that’s the case, I’ll wait for the Destroyer to pass through, and rebuild the town.
- Kazuma, warn them too… Come to think of Kazuma as a terrorist, I’ll show you just how great I can be when I get serious!
- This isn’t the capital of water and hot springs, Alcanretia?
- This way, the gate to the mortal world… Keep this a secret, okay?
- When the guild judges the quest to defeat the Demon King to me.
Short sentences:
- Can you do that, I’ll share some mana with me!
- I might look like a devil!
- You defeated two Demon King’s army is here!
- … Is… Is that so?
- No matter how you look at it, isn’t this a rifle?
Clone the project.
$ git clone https://github.com/MarsRon/aqua
Create a Python virtual environment.
$ python3 -m venv venv
$ source venv/bin/activate
Install libraries.
$ pip install -r requirements.txt
Run the test script to generate a few sentences.
$ python test.py
Start the FastAPI webserver.
$ python main.py
I've included a Jupyter notebook so that you can make your own Aqua.
See train.ipynb
for more information.
Distributed under the MIT License. See LICENSE.md
for more information.
MarsRon - [email protected] - marsron.name.my