Description:

Generating categorized quotes with artificial neural networks.

Inspiration:

Trying to get AI to write novels only produced parts that made sense but did not make sense as a whole so we tried to tone down the scale down.

What it does:

Our project generates quotes when the user selects a category.

How we built it:

We scrapped quotes from Goodreads and then trained the AI on the data on Google Colab. We then tried to port the python neural network model to javascript.

Challenges we ran into:

Removing bad data (quotes in foreign languages). Optimizing data for AI training. Getting categorized quotes. CSS. Porting to Javascript.

Accomplishments that we're proud of:

The "economic" way of generating categorized quotes. Letting the initial string/Seeding the network with the intended category.

What we learned:

Pug (Jade successor). Working within the constraints of an API is easier than trying to add to one (Thinking of different ways to generate categorized quotes)

What's next:

Getting more data. Considering GANs or modifying RNN input.

Built with:

Google Colab: Python, Tensorflow/Keras nodejs, tensorflow.js

Prizes we're going for:

Arteck HB030 Portable Keyboard

Misfit Shine 2

$100 Amazon Gift Cards

Intel® Movidius™ Neural Compute Stick

Team Members

Jinchao Yang, Evan Risas
View on Github