Clipped, an app that automatically generates summaries of news articles using natural language processing (NLP), has released a major update on iOS that creates custom infographics and offers improved accuracy. The startup is also taking sign-ups for an API that will open up its NLP engine to other developers.
With the move, Clipped has dropped official support for Android, but founder Tanay Tandon said he is working with a group of developers to create an open source version for Android.
Similar to the Summly app acquired by Yahoo last year, Clipped uses machine learning to scan an article and then summarize the most important parts. The new version adds a visual aspect by identifying statistics and data points and then creating graphical representations of them.
Clipped version 2.0 also boasts improved accuracy over earlier versions. Tandon claims that the app can now understand major points of an article with 85-87 percent accuracy, up from 70-75 percent. The new app includes a cleaner, more modern design.
When I gave the app a try, the infographics produced mixed results. For percentages and stock shifts, it showed an attractive graphic, but it would sometimes also highlight random numbers, such as the time, in an article. In one case, Clipped got confused by the model number on TNW’s Lumia 930 review.
Tandon started Clipped in 2012 at the young age of 15, much like Summly founder Nick D’Aloisio. Unlike Summly, however, Clipped uses grammatical pattern matching to understand articles, instead of focusing on keywords. Tandon said he built the NLP engine himself with the help of some machine learning libraries.
Moving forward, Clipped plans to expand its API to include other use cases. For instance, the company has been in talks with automobile manufacturers to embed its technology in in-car systems. If integrated, you could have your car read you a summary of the day’s news during your commute home. Tandon also said he’s interested in using Clipped to help summarize non-news content, such as legal and medical documents.
➤Clipped for iOS
Image credit: Paul Tearle