NewsRel Uses Machine Learning To Summarize News Stories And Put Them On A Map by Frederic Lardinois.
From the post:
After 24 hours of staring at their screens, the teams that participated in our Disrupt NY 2013 Hackathon have now finished their projects and are currently presenting them onstage. With more than 160 hacks, there are far too many cool ones to write about, but one that stood out to me was NewsRel, an iPad-based news app that uses machine-learning techniques to understand how news stories relate to one other. The app uses Google Maps as its main interface and automatically decides which location is most appropriate for any given story.
The app currently uses Reuters‘ RSS feed and analyzes the stories, looking for clusters of related stories and then puts them on the map. Say you are looking at a story about the Boston Marathon bombings. The app, of course, will show you a number of news stories about it clustered around Boston, then maybe something about the president’s comments about it from Washington and another article that relates it to the massacre during the Munich Olympics in 1972.
In addition to this, the team built an algorithm that picks the most important sentences from each story to summarize it for you.
No pointers to software, just the news blurb.
But, does raise an interesting possibility.
What if news video streams were tagged with geolocation and type information?
So I could exclude “train hits parade float” stories from several states away, automobile accidents, crime stories and replaces it with substantive commentary from the BBC or Al Jazeera.
Now that would be a video feed worth paying for. Particularly if for a premium it was commercial free.
Freedom from Wolf Blitzer’s whines in disaster areas should come as a free pre-set.
Just a small amount of additional semantics could lead to entirely new markets and delivery systems.