Natural Language Processing, Software Development
Newsadoo - “Spotify for News”
How we leveraged AI to create a customized newsfeed with the goal of bringing the user’s attention back to the world.

The app for a world full of distractions
Newsadoo is an Austrian startup with a remarkable goal: bringing news back into people’s lives, even in a world full of distractions that keep readers knee-deep in entertainment and away from information.
To fit their business proposition, we supported them with Natural Language Processing and Artificial Intelligence – let’s see how.
The Challenge
Tailored, individualized news filtering
Our goal was to use Natural Language Processing and Artificial Intelligence to filter, bundle, and understand incoming news articles. Then, use this information to match the content to the news consumers.
This allows Newsadoo to offer its readers a personalized, multi-platform news timeline in just one service.
Together with Newsadoo, we have built an AI-powered news processing pipeline. Incoming articles are put through data preparation, Natural Language Processing/understanding, and several classification steps.
But let’s get more in-depth.

Collection & Extraction
News can be found all over the Internet. Thousands of websites and other electronic media channels publish updates on millions of topics.
To gather all this information, we developed push-and-pull-based mechanisms that assure that all relevant media coverage is ingested into the Newsadoo news pool. Part of this step is to separate actual stories from advertisements and other interlaced text blocks.
After having the articles in our news pool, we apply different information extraction algorithms to acquire Tags, Locations, Sentiment and more.
Categorization & Bundling
Finding relevant information within this enormous news corpus would be very hard without having categories.
To automatically assign incoming articles to news categories we transform the articles into machine comprehensible vector representations and apply machine learning models to them.
Often there are only a few sources of first-hand information but many news sources covering the situation. This puts users in the uncomfortable situation of having to go through repetitive content -- a problem we mitigate by applying natural language comparison techniques to group articles with similar content, that are then bundled.
Overview & Details
Which do you prefer: short news briefing with important facts, or detailed analysis?
By using models of language information theory, we can present topics in the right level of detail, depending on device and situation.
Are you interested in how Machine Learning / Natural Language Processing / User-Specific Recommendation can be applied to your use-case? Contact us now!

Newsadoo
Newsadoo is an IT startup from Linz that has already won several international awards. Since 2017, they have been building the European alternative to Google News with their news app. Their young team is made of experts in the areas of front & backend development, data science, product marketing, marketing, and partner management.