It’s not easy to find quality news sources on subjects that matter to you. At the TechCrunch Disrupt SF Hackathon this weekend, a team of four strangers came together and using IBM Watson technology were able to hack together an intelligent news reader.
“We used IBM’s Alchemy News API to scrape high-quality news sources based on taxonomies,” Ariel Weingarten, who made the presentation for the team, told TechCrunch. (IBM bought Alchemy for just this kind of use case last March.)
Like many hackathon teams, the four guys, which included two teenagers, were forced to pare down their vision as the night went on. They also had to work around several problems, but they were able to build a usable tool before the night was over.
The team included Weingarten, Joel Ward and 14 year olds Daniel Bessonov and Patrick Li.
The Watson tech is so sophisticated, it can even watch your preferences and begin to use predictive analytics to offer you news and sources that would interest you.
“So many web apps focus on making content available. Very few process the content, directing you through the noise while giving you lots of signals,” Weingarten explained.
But using the Watson API can be challenging and the developers had to work around a number of issues, especially with the short turn-around time of a hackathon.
Among the problems the team encountered was getting stuck trying to build the application in IBM’s Bluemix development environment. The developers felt with enough time, they could have overcome the issues they were having with Bluemix, but given the time constraints, they decided to move into GitHub where they were more comfortable.
They also found in early iterations, they were getting some strange results, but realized it was because the Alchemy API was using Tumblr as a news source. They said when they filtered out Tumblr, they got much more reliable results
They believe that the product could be expanded in a number of ways, such as building on Reddit to get real-time opinion or on top of Wikipedia to get historical information.
What’s more, they could see monetizing the idea by offering subscriptions to users, who are looking for more intelligent news distribution than they can get from from sources like Google News or Flipboard.
Given the time limits, and the level of sophistication of the Watson API, there were lots of parameters they ignored, but if they were to continue expanding this, they would have more time to understand the abilities of Watson and expand the product in a number ways, increasing the sophistication of the news-finding algorithms.