SAMS Forum: Promise of Automating Metadata Remains; Today’s Reality Is Still Unclear

Will computerized, automated logging systems and technologies ever be smart enough to match human operators when it comes to handling the more mundane aspects of metadata logging and tagging? That was the debate during a session at the SVG Sports Asset Management & Storage Forum at New York’s Hilton hotel on July 29.

From left: CBS Sports Network’s Walter Raps, NerVve Technologies’ Thomas Slowe, and ARRIS’s Venu Vasudevan

From left: CBS Sports Network’s Walter Raps, NerVve Technologies’ Thomas Slowe, and ARRIS’s Venu Vasudevan

“In professional sports, it is easy to get data, but what you can’t get from stats is the human aspect,” said CBS Sports Network CTO Walter Raps. “If a player is limping, for example, that aspect can be part of the story around the game. And, no matter how good the taxonomy, it is not going to identify things like that.”

That was the general consensus. However, some opportunities were identified. Dextro co-founder David Luan opined that artificial intelligence has been oversold and is being used to identify things like logos on a uniform to do some of the work in terms of creating categories of content.

“And then the logger can annotate what the segment is about and work on problems that are much more challenging,” he said.

Venu Vasudevan, senior director, multiscreen media and systems, ARRIS, agreed, adding that the goal is to find solutions that can get to about 90% of logging with no errors.

NerVve Technologies CEO/co-founder Thomas Slowe noted that automated systems are ideal for doing things like brand management within a broadcast (NerVve recently inked a deal with Wasserman Media for exactly that function). For example, the system can detect whether a Nike logo is shown on TV.

The challenge is that, in the real world, promising technologies, often fall short of theoretical claims.

Raps cited using closed-captioning data. “It’s supposed to be 90% accurate, but we find it is only 50% accurate,” he explained. “And visualization recognition is fine when players are standing, but, when play starts, it can’t discern the number 81 from 19. There is nothing in the real world that can do the job as well as a logger. And, even if it could detect subjective things like a coach talking to a player, it can’t react as fast as a human being.”

According to Levels Beyond COO Christy King, her company’s Reach Engine platform enables a lot of really interesting automated metadata-input methods that lessen the need for a human touch. “Our platform,” she said, “is about bringing in multiple sources of metadata and layering [them] into the process.”

From left: Source Digital’s Hank Erecon, Levels Beyond COO Christy King, and Dextro’s David Luan

From left: Source Digital’s Hank Erecon, Levels Beyond COO Christy King, and Dextro’s David Luan

And, she added, these systems are about more than simply making content more searchable for production. When the UFC was on PPV in bars and restaurants, data gathered via an infrared light on the set-top box could tell the UFC how many people were in a room and also what kind of events within a fight they found most exciting.

“Also,” King said, “if there is something that happens that has a lot of blood, that might dictate rights issues and what people could do or not do with the content.” Metadata can also help with distribution of rights and signal how long content is available, and, when rights expire, an alert can be sent to the sales team to renew them.

“There are tons of automated processes that can be layered in so people can know when they can do something with footage,” she added.

Source Digital CEO Hank Erecon noted that understanding an audience via things like social-media trending is another form of metadata that can affect a production.

Technology advances like the move to 4K and higher-resolution images may, at first blush, seem a way to better identify players and events, but Slowe pointed out that higher-resolution files create a big problem: processing the larger images takes longer, and that, ultimately, slows down the logging process.

The reality facing the industry is that data is, increasingly, flooding the gates in the form of social media, more in-depth stats, greater analysis of stats, and more. For now, people hold the upper hand in connecting data dots and reshaping productions and rights. But it is fairly clear that machines will eventually play an important part in driving efficiencies that can make those rights and productions more valuable.

Subscribe and Get SVG Newsletters

  • SVG Insider (Tuesday - Friday)
  • SVG Digital Now (Monday)
  • SportSound (Monthly)
  • College (Monthly)
  • Venue Production (Monthly)