Inspiration
Ever crack open your fortune cookie after a delicious meal only to find a fortune you're unable to relate to? Oftentimes the wisdom in fortune cookies goes unheard because the right fortunes aren't able to find their way to the right people. Our project aims to fix that. DeepFortune uses sentiment analysis to accurately gauge how you're feeling on any given day by analyzing your tweets, and delivers you a personalized fortune with exactly the wisdom you want to hear.
What it does
DeepFortune uses the Twitter API to pull the last day's worth of tweets from any Twitter username entered into the system. Then, we take advantage of IBM Watson's tone-analyzer to gauge the sentiments of a user's tweets. We're then able to map any user's combined tweet data to a very specific and personal fortune. A tasteful animation accompanies the fortune on our website.
How we built it
The back end is built with Django and we used the Twitter API to get tweets for any username when prompted. Using Python, we took advantage of the Watson tone-analyzer to generate a series of metrics for the tweet text. By intelligently analyzing and aggregating this data, we are able to algorithmically find a fortune in our database to return. Our frontend was built with HTML, CSS, Javascript, and JQuery. This interactive module allows the user to open their cookie by clicking the mouse to reveal their fortune.
Challenges we ran into
Initially, we spent a significant amount of time trying to auto-generate a larger database of fortunes based on some initial seed of fortunes. We were unable to get the free $100 of AWS credit to work and as a result, we were unable to properly train our neural network in time, since we lacked the appropriate GPU hardware. Bar this, the main challenges we ran into were in data collection, our algorithm for determining the most optimal fortune, and the frontend of our website.
Accomplishments that we're proud of
We think our predictions are fairly accurate, and that our interactive web module is innovative and unique. There could also be a significant future application for our project, in that restaurants could implement similar systems to enhance their customer experience with personalized fortunes.
What we learned
Deep Learning requires adequate hardware and Amazon Web Services has a steep learning curve.
What's next for fortune-cookie
Eventually, we'd like to implement a system where we can train a neural network with a number of tweets that correlate to certain fortunes, such that our system can intelligently write its own fortunes based on a user's tweet.
Log in or sign up for Devpost to join the conversation.