Inspiration
Twitter is littered with opinions from all political sides but it can be extremely tiring to get tweets about a topic from sides to get an overall picture. That's where we come in. Enter a search term like #CapitolRiots or #Biden46 and our trusty NLP model will return a couple lists of segregated tweets about the topic.
Feel free to use our attribute search and get the most popular tweets, tweets from verified accounts, or in a specific date range.
See a tweet and wanna see what side it belongs to? Or maybe a tweet has a hashtag that you'd like to use the website for? No need to bookmark our website, simply @Unbiased_bot in the replies and the bot will get back to you with an answer and a direct link to the website with the hashtag.
What it does
How we built it
Flask - Python's trusty backend for small applications HTML - you know what this does CSS - Clean and simple UI which makes the application responsive and good on phones Bootstrap - great UI boilerplate for modern web apps JavaScript and jQuery - adds to the backend AJAX - takes care of the async forms on the tweet page Scikit-Learn - Training and Testing the NLP model Google's BERT - Pretrained weights for quick NLP training Firebase - host realtime database for taking user's input and for training the model concurrently with new data Tweepy - Twitter API's python wrapper for the bot and the app Heroku - Host the application online
Challenges we ran into
AJAX was tough to learn but thankfully we didn't run into a lot of problems.
Accomplishments that we're proud of
It works, phew. The model can be more accurate but to even have an app with an NLP model built into it was pretty cool for us.
What we learned
Putting together so many technologies can be hard.
What's next for Twitter, but unbiased
Improving the model, making the bot more and more accurate, and hoping that people can actually use it.
Log in or sign up for Devpost to join the conversation.