Bringing Together Data Scientists, Developers, and Journalists for Hacking Journalism: Data Science
Before I begin, just check out the sounds here. This is wonderful.
Over the last few months Embedly has been working on a publisher product built on top of the video analytics. Analytics can take time to parse - could there be a tool informed by those numbers that could offer clear actions and value? That’s how Recommend came to be - video recommendations based on what people are finding interesting in real time.
Those steps from audience behavior to product can be a powerful path. What other strategies are there from data to product? What other tools can be built in the publisher space with this methodology? The last year has seen growth in the data science space in newsrooms. It’s no secret that BuzzFeed’s success is in part because of their strong data science team.
Making a Space to Share Ideas Around Data Science in the Newsroom
Embedly teamed up with the Washington Post to produce the third Hacking Journalism, on Data Science. After Bezos bought the Washington Post, a data science team began - propelling the news org to the leading edge with their tech, news, and brand.
Data scientists, developers, journalists, and designers gathered at the Washington Post to share ideas and build prototypes. Participants came from a range of media companies, including AJ+, AOL, Gannett, Hearst Newspapers, National Geographic, NBC Universal, NPR, PBS NewsHour, Pulitzer Center, The New York Times, The Wall Street Journal, The Washington Post, Turner Media, USA TODAY, and Vox Media.
At the beginning of the event there were talks, including one by Rohit, the Embedly Data Scientist, on recommendation systems. People enjoyed the variety of possible recommenders, including the dislike recommender.
Kawandeep Virdee on Twitter
Sean gave an afternoon workshop:
Kawandeep Virdee on Twitter
Sam Han leads the data science team at the Washington Post. He gave a powerful talk on what they're working on, including their virality prediction work.
Kawandeep Virdee on Twitter
The Washington Post produced video interviews and converage on the event:
The Washington Post hosts Hacking Journalism event
Here are a couple writeups of the event:
Thirteen cool ideas from the Washington Post/Embedly hackathon - Storybench Hacking journalism: Data science in the newsroom - 3 to read
The emphasis of Hacking Journalism was to build relationships and learn new techniques. While some of the projects may go on to production in newsrooms, they should be seen as prototypes to start a dialogue. If you are interested in any of the ideas, reach out to the creators.
DCSN - Optimize calls to action (like share buttons) based on the reader context, including the referrer and subscription status.
RenBroc - A method to identify engaging content that otherwise has received little attention. It surfaces articles that have a lot of comments, but lower than expected pageviews.
Full Story - Given a news story, this tool presents snippets from various articles presenting that story to highlight differences in opinion and framing.
Quotable - Browser extension that lets you highlight and share snippets of articles. These highlights are used to inform an article recommender, which includes paid articles.
Dig.it - Surface archived articles relevant to current topics. Determine keywords of trends by using tweets of trending hashtags, and search through the archives using these keywords.
Flinder // Tinboard - Surface content to the reader based on how much time they have to read.
Audiolization - Easily represent data with sound.
Journofriend - Choose a subject and get updates from a variety of sources on it using topic analysis and categorization
The Bridge - Analyze a given list of twitter users, and observe trending hashtags in their conversations.
Content Magical System - Write, edit, and publish articles through e-mail. Designed for publishers with a dashboard to track work.
Make Me Smart - This dashboard gives press secretaries feeds that include national news, tweets, geographic trends, and government reports around a given issue.
oldNews - Timehop meets article recommendation. Get recent article suggestions given what you’ve read a year before or earlier.
File This - Bookmarklet that saves and indexes web pages, extracting keywords and entities for filtering and search.
David Rouse on Twitter Jessica Morrison on Twitter Aleszu Bajak on Twitter Kawandeep Virdee on Twitter bmuller on Twitter Kurt Yalcin on Twitter
You can find more information on all of the projects here:
HackingJournalism Data Science - HackDash
Hacking Journalism: Data Science was possible with the organizing team: Laura Bucci, Angela Wong, Alex Remington, Leonard Bogdonoff, Lam Thuy Vo, Jeanne Brooks, Matt Carroll, myself, the support of the following companies: Embedly, Washington Post, Conde Nast, Mozilla Open News, Chartbeat, MIT Future of News, Elastic, and Huge, and additional help from Jane Lockhart and Carlos Coral from the Washington Post.
Leonard Bogdonoff on Twitter
Also my parents came to the event which was the best.