Mindless is a fun web application built using an N-gram language model. In particular, we choose 3-gram because it gives the best performance. We use Chicken Soup for the Soul as the training corpus to generate word and context matrices. Based on Markov assumption, we continuously generate the next word using the probability calculated from the word and context matrices. We also use add-k smoothing to overcome the zero-count issue.
Mindless can help people de-stress. It's a fast-paced society, and we sometimes get anxious and overwhelmed by all those to-dos and deadlines. Through looking at a cute character talking gibberish, we derive fun, happiness, and inspirations to those unanswered questions in our heart.
Mind LESS and laugh MORE with Mindless 😉
You can try out Mindless at https://mind-less.herokuapp.com.
You can use Python 3 and Node.js to build and run the project.
Install the Python dependencies using
pip3 install -r requirements.txt
To compile the frontend assets,
cd web
yarn install
yarn build
To start the server,
python3 app.py
This spins up a Python Flask server that servers the frontend asset,
and also provides the API to the backend language model via /api/talk
There's some rough edges that hasn't yet to be evened out. For now,
comment out the front-end build folder from the .dockerignore file
so that it'll be included in the built image.
Here are some of the examples:
