candidats.media came from the theory that today, when an election occurs, citizens mostly vote for the candidates that are the most present in the media.
To verify that theory, we started analyzing the news of the french political press, extracting content from 24 different news outlets, and calculating the presence of every candidate in the written online media. Along that analysis, we thought it would be interesting to know if journalists are speaking from these candidates in a positive or negative way.
The website is still online, even if we stopped the analysis after the election of the new french president, Emmanuel Macron, on the April 23, 2017.
FEWS (for Fastest NEWS) was a project dedicated to saving time to readers of the online Tech press. Something that was annoying each one of us was to have to follow the major Tech blogs/websites to get all the news, and then have to read several times the same article while going through the aggregated feed of all these news. We therefore decided to automatically detect and put together articles speaking of the same subject. Interesting enough, with the text of all these articles, it was also fairly easy to generate a 200 words summary automatically. The rule was, that the most article are speaking about it, the higher it stands in the list.
For this project, we analysed 30 different tech blogs/websites.
YoNews was our first successful experiment with data extraction and Natural Language Processing (NLP). Yo organised at the time a challenge around its' platform, to generate applications working on top of the simplest communication tool ever created: sending YOs. Our take on it was YoNews, a set of channels created around the big themes in Tech News (Apple, Google, Apple, etc.). When "Yo-ing" one of these channels, the channel would give you the "most relevant" article speaking about that keyword, and it would never send you the same article twice.
The Youtube promo video is still online.