How AI transcriptions have transformed audio content for PRISA Media
With more than 50 brands, across news, sports, music and entertainment, PRISA Media is the leading Spanish-language media group in the world. Some of their brands are niche, others are national, and some, like El PAÍS and the sport newspaper As, are global. PRISA operates newspapers, radio stations and podcasts in Spain, Chile, Colombia and Mexico.
PRISA has been using AI across a broad range of activities, including coding and advanced analytics, Diego J Pruneda Paz, the publisher’s Director of Transformation, Innovation, and Voice Projects, told participants at WAN-IFRA’s Frankfurt AI Forum this week.
They’ve also used AI to develop a fact-checking tool for audio to identify deepfakes. Other use cases for AI include several initiatives relating to content generation and repurposing, as well as customisation and recommendation.
PRISA has three main areas where they look for opportunities for using AI, Pruneda Paz said.
The first is to help unlock more value from their vast content catalog.
The second is efficiency. “AI can help us accelerate the workflows, and bring down costs. It can also make feasible things that were impossible to do before AI, so we actively look for opportunities here,” Pruneda Paz said.
The third area is differentiation, meaning initiatives that help show what makes PRISA different from other outlets and from platforms, he said.
Boosting possibilities to repurpose content
These opportunities are especially important for PRISA around audio.
“We run one of the largest audio operations in the world with more than 1,200 journalists on radio stations that are broadcasting live 24/7. Our newsrooms range from very big national hubs to tiny newsrooms with less than five people, so we run a complex and diverse multi-format and multi-country operation,” Pruneda Paz said.
Unfortunately, audio has largely been a sort of “black box” that has been ignored by traditional search engines, which he noted are almost always optimised for text rather than audio.
Because of this, PRISA sought to create AI solutions to improve accessibility and use of both the audio content their journalists are creating as well as their vast audio archives.
For example, PRISA uses AI to transcribe all of its audio, which is then input into a multimedia system that their journalists can search for transcripts. They can extract quotes easily, and they can create summaries of those transcripts, which can help create articles for online or used for social media.
PRISA received a grant from the Spanish government for the adoption of AI initiatives within media companies, which made their project viable, Pruneda Paz said.
“At its core is a set of AI-powered processes that let us generate, transform, search and connect our content,” Pruneda Paz said. “We chose content from several different formats, from several different content management systems, and then build services that expose those capabilities to any CMS or product that PRISA has,” he added.
All audio is now transcribed into text, and the system also offers transcripts as subtitles, so journalists can get the audio, the transcript and an image, and create a short-form video, for example.
“Even though transcriptions are not a very fancy technology, before now, people had to rely on standalone solutions where you had to have licenses for everybody, and we have several hundred different newsrooms,” Pruneda Paz said. “With this technology, transcriptions are everywhere and this alone changed dramatically how audio is used in the newsroom.”
