The Tinder interface is catching on all over the place. That satisfying finger swipe is no longer confined to images of people you don’t want to date; it can be used for shopping and all manner of other sorting tasks, and thanks to a new app called Newsly, people can now set their swiping fingers to work on sorting their preferred news content.
The app was built overnight by three University College London students, who used machine learning APIs to build a news reader that learns how to pull up a user’s personal news preferences. A card appears on screen containing the article image, headline and sub-headline. If it’s not something you’re interested in, a quick swipe to the right will dispatch it in a blink (like Tinder).
The app learns to tailor the articles it pulls to the users’ preferences. About twenty swipes is enough to give the algorithm enough data to catch the drift of your interests, but cleverly, it also learns what brings on your yawns of indifference. One of its creators, Davit Buniatyan, told TechCrunch journalist Natasha Lomas:
“The Facebook News Feed is based on 20 previous likes. So it’s kind of this one but the advantage over Facebook is we have also dislikes which is very important in training [the algorithm]. Every time you swipe it will learn more about you.”
Buniatyan is clearly a commercially astute young man: the user data the app generates, he said, will have “clear value for advertisers wanting to target ads.”
Newsly uses the Intel Mashery API, which pulls stories from the online version of The Guardian newspaper. Other news sources can be added relatively easily to broaden the base of news stories as well.
Buniatyan’s UCL colleague, Muhammad Rafdi, explained:
“It searches through the [news] content — it does a text analysis, and the text analysis will produce keywords which are then stored into the server. After several queries it builds up a new one and then that will fetch a new list of news feed based on your preferences.”
Although the idea was derived from Tinder, they implemented their own algorithm, Nafdi added. And to ensure that readers’ world views don’t narrow down too far, the app resets preferences after a specified time has elapsed.
That “Tinder + Machine Learning” mix could go far.
