Seminar: Natural Language Generation, Automated Journalism and Finding the Middle Road

Welcome to all!

TITLE: Natural Language Generation, Automated Journalism and Finding the Middle Road
WHEN: Friday, 24 September 2021, at 11.00-12.00
WHERE: https://uib.zoom.us/j/63314203494?pwd=dGM0MlkwR1QwUjY5RW9UZXR1VnJ3QT09#success
Meeting ID: 633 1420 3494
Password: TC17bnzj

After more than half a century of academic research, natural language generation has established a clear foothold in the newsroom. In this talk, I’ll discuss lessons learned from building multilingual natural language generation systems for several number-heavy news domains such as elections, national statistics, and COVID developments. We’ll discuss what I believe are the most significant relative downsides associated with commonly employed approaches, and what could be a “middle road” that might sidestep at least some of said downsides. This exploration is not limited to purely technical aspects, and we will also make short detours to some legal and ethical questions related to automated production of news texts.

Leo Leppänen is a computer science doctoral student at the Discovery Research Group of University of Helsinki, Finland. His research focus is on data-to-text natural language generation, more specifically on the automated generation of reports for purposes of news, analysis of historical news paper collections, teaching etc. In addition to approaching the topic from a computer science perspective, he is also interested in both ethical and legal questions related to use of AI in the newsroom.