Our idea is inspired from John Oliver's show which has touched upon many social issues from time to time. We were inspired by this to create personalized newsfeed timelines of social and local issues. Our aim is to always keep people aware of the social issues around them.

What it does

Imagine a simple web application and email service that keeps you updated about the social and local issues you care about. IssueMail is for people who care about social and local issues, but are too busy or don't have a unified platform to do so. People who sign up can choose which social and local issues they would like to follow and IssueMail gives them with a timeline of articles based on the issues they followed. With the sign up the users can also subscribe to periodic e-mail service which keeps the user up to date with the new articles as well as their timelines.

How we built it

Our application had different components - Data Fetching, Machine Learning and Ranking, Web Server and UI.

  • For fetching data, we use open apis for news articles, like guardian and alchemy

  • For Machine Learning, we used sckit-learn python package, and Alchemy Sentiment to get sentiment of posted articles

  • For Web Server, we run a Flask web app on IBM Bluemix Cloud Foundry container.

  • The frontend is powered by AngularJS, backed by the Firebase authentication library.

Challenges we ran into

Due to the API restrictions we were restricted to a relatively smaller data-set to test out machine learning algorithms. Also, due to other API restrictions we decided to use just one source of articles for demonstration.

Key Features

  • Timeline of Articles - For each issue we show the timelines of the best articles.

  • Personalization - User has the ability to personalize the feeds of timelines based on issues they follow.

  • Sentimentality of Articles - IssueMail uses IBM Sentimentality analysis to classify articles as either positive, neutral or negative.

Accomplishments that we're proud of

We were able to succesffuly have a service which grabs relevant content, applies Machine Learning algorithms, and has a usable graphic interface. This looks a product ready to ship!

What we have learned

  • Identifying the correct News API. It's essential to work with an API that provides comprehensive coverage over dates to have the data we would have loved to have

  • There's no perfect clustering algorithm to deal with such vast vocabulary. We currently stuck to soft K-means, however, extensions to our algorithm like LDA can be integrated in future.

What's next for IssueMail

For future we would like to expand the list of social or local issues or let the user decide them. We would also like to partner up with some NGO's which may be looking for people to volunteer and use the platform to get them connected.

Share this project: