WATCH THE FULL VERSION OF THIS MASTERCLASS NOW
NEAL ROMANEK: In the world of news, your challenge is less about big media files and more getting content out at speed. What benefits can AI bring your newsroom customers?
areas. It increases productivity by automating some of the tasks. It helps to index content, whether it’s videos, audio or pictures. And it aids finding content. Like Bill said, we have customers managing libraries with millions of assets. It’s hard to find content, especially when the metadata is not organised. Speech to text is probably the most common technology that we’ve explored. With it, you can index content, but also generate a first version of your captions using a combination of natural language processing and speech to text. It’s also about the capability to fix or improve captions, and make content compliant with broadcast standards. Facial recognition, especially when you have a large amount of content to index, is also a useful tool. The idea is to analyse the archive and find common faces, then either automatically or manually tag them. That can improve archive documentation. We see more custom-made models too, for specific things
MICHAEL PFITZNER: A newsroom system is all about
into the system. And we’re working on other parts – like text summary, abstract building and keyword extraction. If you have a technology that can analyse text and content, there are tons of possibilities of how to use the results, regrouping and re-offering them to our newsroom customers – to create a big value and help with these tasks that couldn’t be done by humans. We have also talked about evaluating the diversity aspects of broadcast stories here, or the balance between women and men, for example.
metadata. Using our software, we integrate semantic search and topic extractions. With the sheer volume from agency wires, there must be mechanisms that help you aggregate or set a cluster around topics. And this is already done by AI. We can use this functionality to presort and pre-analyse the incoming sources. There are also mechanisms we can use to do a fake news detection ahead of time, and give teams a reliability rating alongside the ingested text or other material. Things like speech to text make it easier for journalists to bring stories
RAOUL COSPEN: We have found that AI helps in three different
“IF YOU HAVE A TECHNOLOGY THAT CAN ANALYSE TEXT AND CONTENT, THERE ARE TONS OF POSSIBILITIES OF HOW TO USETHE RESULTS”
WATCH THIS VIDEO WITH QR CODE
– like a company wants to recognise a sponsor’s logo. They can create a model for the logos that is able to say whether it’s partly or fully visible, or if it’s blurred.
MICHAEL PFITZNER, CGI
feedmagazine.tv
Powered by FlippingBook