SEA XL: Algorithms and the News


This XL version of SEA features talks on "Algorithms and the News".

09:00 Doors open
09:30 Anne Schuth - Personalized Regional Push Notifications
10:00 Daan Odijk - Your News, Our News
10:30 Dung Manh Chu - Mining Knowledge in News for Personalization
11:00 Edgar Meij - AI & Automated News

Night Room, Startup Village, Amsterdam Science Park


-- Anne Schuth (Team Lead/Machine Learning Engineer, De Persgroep)
-- Personalized Regional Push Notifications
-- Every day, journalists at the Algemeen Dagblad write over 500 news articles. Just a fraction are currently sent as push notification to our readers. Human editors decide whether an article should be pushed to millions of our readers. They have no means of targeting small subsets of our readers and therefore only maximally send out 10 push notifications a day. We are changing this process by personalizing what we push. The end goal is pushing every article that is published to at least some users that would find it interesting. In order to do so, we need to know as much as we can about these news articles and our readers. We apply natural language processing and we analyze reading behavior. We focus on pinning our content and users on a map of the Netherlands so that we can send push notifications for content that happens in regions the user is interested in.

-- Daan Odijk (Lead Data Scientist, RTL)
-- Your News, Our News
-- The mission of RTL Nieuws is to inform its viewers and readers in an independent and accessible manner about important developments in the Netherlands and the rest of the world. The news service produces several daily news bulletins on television for RTL4 and RTLZ, reaching a total audience of about 1.5 million people. Furthermore, through the RTL Nieuws website, apps, and social channels, over a million users are reached every day. I will share how we are using search technology to help our news readers find the right content for them. I will also present two upcoming projects: one on measuring diversity in news and another on automatically generating local news articles.

-- Dung Manh Chu (Data Scientist, FD Mediagroep)
-- Mining Knowledge in News for Personalization
-- The FD Mediagroep is the primary source of financial economic news in the Netherlands. Serving a wide variety of users with different backgrounds, interests, and contexts, we believe that serving our digital media (text and audio) in a personalized manner can be beneficial. In this presentation we present our building blocks in achieving this final goal: personalized newsletter with collaborative filtering, entity linker to companies, sentiment analysis and BNR tagger. We show that with rich data, simple AI models have been performing well in our cases.

-- Edgar Meij (Team Leader and Senior AI Scientist, Bloomberg L.P.)
-- AI & Automated News
-- While the technology is far from mature, artificial intelligence in the form of autonomous production of journalistic content is becoming increasingly prominent in newsrooms – and it’s here to stay. The promise of automatically generating news at a faster pace, a larger scale, in multiple languages, and with potentially fewer errors, has scholars and practitioners championing this technology. This development fuels fears that journalists will soon be out of work. Yet, today’s algorithms cannot ask questions, explain phenomena, or establish causality, giving human journalists the opportunity to write stories that address the ‘why’ – as opposed to the ‘what’ that machines tell us. When established news organizations start publishing partly or fully automated news stories, they lend credibility to them. Little is known about societal implications of this on trust and potential bias, as the algorithms themselves cannot be held accountable. I will discuss these developments and place them in the context of news search and recommendation, automatic media monitoring, polarity detection and sentiment analysis.