BluCheck

Political tweet sentiment analysis using Co:Here sentiment analysis technology.

Comment

Inspiration

This project is inspired by Co:Here's challenge to create a trained sentiment and language processing application. It is also inspired by the Berggruen Institutes Challenge to create a product that facilitates deliberation in a community by presenting current political trends in tweets. We wanted to create a product with a positive impact that could help recognize racism and extremism on Twitter. We believe that data must be accessible for governments and institutions to address online hate speech.

What it does

BluCheck has been trained with thousands of examples of regular tweets as well as tweets marked as containing hate speech. It is able to flag hate speech in tweets and read a stream of incoming tweets to establish trends. It has also been used to parse large datasets and find trends in hashtags. For example, it found the most commonly used hashtags for democrats and republican politicians.

How we built it

BluCheck was built using python and the Twitter API Tweepy. Co:Here technology was used to train the AI.

Challenges we ran into

Our team lacks experience in front-end development. As such, in the later stages of our product, we struggled to build a website. Also, the AI's sentiment recognition can be quite inaccurate which we struggled with fixing. For example, it would mark any tweet with the word "Waffle" in it as hate speech.

Accomplishments that we're proud of

All of our team members are hackathon beginners. We are very proud of the dedication we showed over the weekend and we worked all night. We are also proud of the fact that we attempted to use AI for the first time. This may lead to exciting possibilities in our future.

What we learned

  • How to use APIs to create more interesting projects.
  • How to train an AI model.
  • How to parse large datasets.
  • Software development practices in python.

What's next for BluCheck

BlueCheck lacks a fully functioning front end which would be our next priority. Afterward, we could train our AI model with a larger dataset to get a more valid sentiment analysis. We believe that BluCheck is powerful enough to one day be very useful.

Built With

Share this project:

Updates