Inspiration
Parents get fairly tired, so we wanted to take away some of the burden of having to come up with a story after a long day of work
What it does
Prints out a new children's story based on children's stories it has read
How we built it
We trained a neural network for text generation on parts of a massive data set of children's stories we got from Facebook's own neural network research
Challenges we ran into
We didn't have enough time to fully train it on as large a data set as we would have liked
Accomplishments that we're proud of
It seems to be functionally learning things
What we learned
A lot about machine learning, which along with python and anything in google cloud services nobody in our group had done before. We also learned that Google lets you do machine learning on their virtual machines up to a certain amount, which saved us hours of waiting for my computer to process it.
What's next for Bedtime Stories
We'd like to spend a lot more time training it, and add a library for sound so it can read them out
Built With
- google-calaboratory
- google-cloud
- neural-networks
- python
- python-bottle
Log in or sign up for Devpost to join the conversation.