Building knowledge graphs with media content and application for tackling misinformation

The sheer volume of multimedia content created every day across the many disparate distribution channels keeps challenging any traditional

content management system. Routine tasks such as archiving, editing,

content organization and retrieval by multimedia creators become

prohibitively costly. On the user side, the amount of multimedia content

pumped daily can be simply overwhelming and the need for shorter and more personalized content has never been more pronounced. Recommending, enriching and summarizing content can help to capitalize on users’ engagement and generate their interactions.



In this talk, I will first present several Knowledge Graphs developed

with medias being radio and TV programs broadcaster in Europe, news items published by a news agency or even social media posts shared on Twitter or Facebook. Next, I will present a number of tools that contribute to automatic multimedia understanding by computers, ranging from extracting topics using common sense knowledge or language models to recognizing recurring people in images. I will show how difficult it is for computers to decompose content into meaningful segments. I will show how to extract highlights from media content, both for narrative-focused summarization and for maximising memorability and I will conclude with our current efforts aiming at tackling misinformation.