The story is simple. I've very ignorant when it comes to my understanding of Politics, and thus, it is very hard for me to know what my friends' political leanings are. This is where Polly comes in. Polly provides an easy mechanism that is both inconspicuous (everyone has a phone), so it's not super suspicious. and easy to use.
What it does
It's a Google Home app built using a self-trained NLP model to recognize certain inputs. Afterwards, we take those inputs and run them through three separate Azure APIs, all of which are cognitive services. We run them through mood, keyword, and sentiment. We then use that data to attempt to extrapolate what the comments' political leaning is.
How we built it
We built it using NodeJS to interface with our NLP model, which we built using google's provided SDKs. We then built upon that a Flask server to serve requests and send data to our google home app. We then made the Azure API calls as the user request was coming in, as well as ran the data against our model to predict what kind of political leaning the statement had.
Challenges we ran into
Creating the model was quite difficult, but tensorflow is very good at doing it.
Accomplishments that we're proud of
We're pretty proud of how much work we'd been able to do in such a short span of time.
What we learned
How to Tensorflow
What's next for Polly
We see Polly as an app that can be utilized to spread knowledge about public policy choices. There are just way too many people who don't understand policy decisions and politics in general, and educating them is a great way for the American people to make an informed decision come election season.