Made at YHack 2017.
This application was designed to prevent the spread of fake news throughout global and domestic policy for the United States of America.
There has never been a more relevant time in political history for technology to shape our discourse. Clara AI can help you understand what you're reading, giving you political classification and sentiment analysis so you understand the bias in your news.
Clara searches for news on an inputted subject and classifies its political leaning and sentiment. She can accept voice commands through our web application, searching for political news on a given topic, and if further prompted, can give political and sentiment analysis. With 88% accuracy on our test set, Clara is nearly perfect at predicting political leaning. She was trained using random forest and many hours of manual classification. Clara gives sentiment scores with the help of IBM Watson and Google Sentiment Analysis APIs.
We built a fundamental technology using a plethora of Google Cloud Services on the backend, trained a classifier to identify political leanings, and then created multiple channels for users to interact with the insight generated by our algorithms.
For our backend, we used Flask + Google Firebase. Within Flask, we used the Google Search Engine API, Google Web Search API, Google Vision API, and Sklearn to conduct analysis on the news source inputted by the user.
For our web app we used React + Google Cloud Speech Recognition API (the app responds to voice commands). We also deployed a Facebook Messenger bot, as many of our users find their news on Facebook.
Lack of wifi was the biggest, putting together all of our APIs, training our ML algorithm neural network, and deciding on a platform for interaction.
We've created something really meaningful that can actually classify news. We're proud of the work we put in and our persistence through many caffeinated hours. We can't wait to show our project to others who are interested in learning more about their news!
How to integrate Google APIs into our Flask backend, and how to work with speech capability.
We want to improve upon the application by properly distributing it to the right channels. One of our team members is part of a group of students at UC Berkeley that builds these types of apps for fun, including BotCheck.Me and Newsbot. We plan to continue this work with them.
react-native - flask - firebase - google-cloud - google-search-engine - google-vision - ibm-watson - google-cloud-search - google-web-speech-api - sklearn -
