The Many Paths on the Road to Personalization

image of woman AI facial recognition
(Image credit: Cr: Witthaya Prasongsin/Getty Images Cr: Witthaya Prasongsin/Getty Images)

NAB Show once again is helping attendees push the creative envelope with focus on content personalization, automated production and using data to engage viewers more deeply in entertainment, all of which fall under the umbrella of Intelligent Content being highlighted in the Intelligent Content Experiential Zone in the West Hall.

It’s easy to understand why intelligent content is turning heads. With the assistance of the cloud and artificial intelligence/machine learning (AI/ML) tools, broadcasters, producers and others are finding new paths to creativity and more opportunities to monetize content.

“AI for intelligent search is becoming almost table stakes for content creators across the board,” said Sean King, senior vice president and general manager for Media, Entertainment and Advertising at Veritone.

“Regardless of the media type, these groups are creating so much content on a daily basis and have such large content libraries that having the ability to find what’s needed gets harder and harder without AI. It’s all about extracting value from content.”

Personalization

Whether it’s a recommendation engine that presents viewers with possible content of interest, a TV reporter trying to find a specific piece of raw or stored footage or a documentarian wishing to buy historical footage, the gas in the engine of these searches is enriched metadata.

“The key — the very first thing — to [enabling] personalization is to enrich the metadata, whether it’s on-demand or live,” said Julien Signes, executive vice president and general manager of Video Network at Synamedia.

An AI natural language engine can extract meaningful metadata for each frame of video from speech, text and subtitles. Similarly, AI tools enabling facial and object recognition can create relevant metadata for each frame. Together, these and other AI algorithms bring a new level of precision and visibility into content repositories, said Signes.

“You need to have a rich data set,” he said. “Otherwise, AI is useless.”

The good news for M&E companies is that “search” and “recommendation” are among the more mature AI-based media applications. “Some areas of M&E are advancing more quickly than others in adopting cloud-based workflows,” said Francesco Venturini, corporate vice president, Communications Media Industry at Microsoft. “For example, data analytics and recommendation engines are more mature and widely used in cloud environments, as they leverage big data and AI capabilities to deliver personalized and engaging content to customers.”

TV News

While personalization is often thought of in the context of SVOD subscribers and recommendation engines, it also pertains to television news — whether it’s a reporter searching a MAM for specific historical footage, a news director looking to review the work of specific reporters on a given topic or a news producer looking to unearth societal trends to put stories into better context.

“AI is able to source content automatically, based on a search,” said Philippe Petitpont, CEO and co-founder of Newsbridge. “A reporter can say, ‘I want a baseball stadium at night in California,’ and automatically AI can search footage and show in the search engine, relevant content that will be useful. In a second, the reporter can have all the images, and that can be done over a thousand, over a million hours of video.”

In the news marketplace, speed to air is one important element of staying competitive — another way AI is well-suited to the newsroom.

“AI speeds up the discovery of raw material,” said Paul Shen, CEO and founder of TVU Networks.

“In the past, you could spend two hours to produce a story. Now, you can do that within minutes. Not only can AI discover content, but it also can be used to feed reporters clips relevant to their stories.”

Helping reporters create stories extends beyond finding relevant content in an instant, however.

“We’re in a place where you could leverage existing AI models and do a first pass on the creative side, writing a story comparing, for instance, today’s banking world to the mortgage crisis of 2008,” said Veritone’s King. “You will still want a human in the loop, but this can really accelerate the process.”

AI also will play a role in personalizing stories for distribution on various social media platforms — not simply conforming to aspect ratio requirements but best practices when it comes to story length.

“A generative AI [artificial intelligence that generates content, such as ChatGPT,] workflow can take an existing story and generate a 30- or 60-second summary [for distribution via social media],” said King.

Monetization

One important capability of NextGen TV is the ability to support content personalization. With ATSC 3.0’s internet back channel from smart TVs and other devices, two-way connectivity with viewers can happen, bringing personalization to viewers and new revenue to broadcasters. The cloud will likely be an important enabler of this personalization, and AI could be used to unearth viewer preferences to enable better targeting of ads and content.

“ATSC 3.0 is a segmented broadcast/IP hybrid approach, where personalized ads or content can be downloaded over the consumer’s internet connection and then triggered by ad insertion markers in the main video stream. This is akin to how OTT providers use SCTE-35 to signal an opportunity for an ad server to deliver an ad for that user,” said Evan Statton, senior principal architect, M&E, at Amazon Web Services (AWS).

“In the case of ATSC 3.0, the broadcaster or playback client would likewise maintain information about a viewer, then use that data to query the ad server to determine the best advertisement for that user. The advertisement would then either be scheduled into the main output via the ATSC 3.0 non-real-time (NRT) subsystem or delivered via the internet to the viewer’s device.”

Beyond commercials delivered to individuals based on their interests, personalization could drive new program packaging options for linear broadcast. One centers on marrying recommendation engine tech with the Electronic Program Guide (EPG) in a 3.0 world.

“[There’s interest in a] personalized EPG experience. This would prioritize the most relevant and interesting [programs] to you [across linear TV channels] where they might be ranked from one to 30, and your one through 30 will be different from mine,” said Greg Riker, chief revenue officer at ThinkAnalytic.

While exceedingly capable for a variety of applications, AI has an important limitation, said Synamedia’s Signes. “Don’t assume AI is a magic wand,” he said. “It’s something that needs to be [deployed on] a case-by-case basis and with patience. Over time it improves; it gets better progressively; so be patient.”

Copyright NAB 2023.

Phil Kurz

Phil Kurz is a contributing editor to TV Tech. He has written about TV and video technology for more than 30 years and served as editor of three leading industry magazines. He earned a Bachelor of Journalism and a Master’s Degree in Journalism from the University of Missouri-Columbia School of Journalism.