Munch.

• 

  Munch Website

• 

  Munch Logo

Inspiration

No one can sleep without a bedtime story, but aren't there a finite number of fairy tales in the world? Not anymore.

What it does

Need a story? That's what Munch is for! It autogenerates stories for all your bedtime or fairytale needs! You can even add buzzwords for your story to feature specific elements (Like princesses or unicorns!)

How we built it

Using TensorFlow, we built a long short-term memory recurrent neural network model that analyses large text files of fairytales and created a durable model. We then utilized Django to create a robust Python based backend for our web product. The entire front-end design is a customized HTML, CSS and Javascript product that makes for sleek interactions with the user. The dataset we used was retrieved from a facebook database.

Challenges we ran into

We had limited time and computing power so we had to wait about 10 hours for our model to be trained to a satisfactory level. This is the first time any of our team members have ever done neural network training on text datasets. This is the first time some of our members worked with web development products, such as Django and Javascript, so it was a challenge working with systems we were not familiar with.

Accomplishments that we're proud of

It works. It produces coherent sentences which are grammatically correct. The website's design also manages to host all our intended functionality. We think the design looks good too.

What we learned

HTML, CSS, Javascript, Machine Learning, TensorFlow, LSTM Recurrent neural networks. Text processing. And implementing all together.

What's next for Munch.

Training it longer and on more datasets. We also want to observe the relationship between text and story telling patterns to have more creative and interesting stories.

Built With

• css
• django
• html5
• javascript
• python
• tensorflow