Improving Reuters stories with exclusive grassroots media fetched in real-time from social media

What it does

It allows monitoring Twitter for user-posted, likely original media, as soon as the first story is there. The app may, therefore, fetch exclusive user-generated content from the epicenter of high-impact events like natural and man-made disasters, political and sports events, etc. The app generates a live feed of fetched images to be accepted or declined by an images curator.

How we built it

We integrated all APIs in Python: Reuters' news feed subscription, Reuter's entity tagging to find location, time and event type data in the article's text, Twitter's search API to find relevant images and Microsoft's Vision API to check if those images are relevant. Then we present pictures using simple web application written in Django and JQuery.

Challenges we ran into

Limitations of APIs. Lack of embedded precise metadata, like exact geolocation of events.

Accomplishments that we're proud of

Working proof of concept. The workaround with MS Azure cognitive services API.

What we learned

Using MS Cognitive services API, Twitter's and Reuter's APIs;

What's next for Newsbee

Automation of interaction with user improving precision of tag matching

Built With

Share this project: