Neither Bryan Cranston nor Kevin Spacey is likely to be at the Bellagio Hotel in Las Vegas on Jan. 8 when the National Academy of Television Arts & Sciences presents the Technology and Engineering Emmy Awards.
The lack of star power, however, doesn’t detract from the importance of the awards. Indeed, the award for “Personalized Recommendation Engines for Video Discovery (PREVD) for MVPDs” is particularly important: No matter how riveting the exploits of Walter White or Frank Underwood, they don’t matter much of nobody can watch them.
The cable industry has waged a long and interesting struggle to get its arms around a simple question: How does it get subscribers to the programming they want, before they know precisely what that programming is? This has been a problem since the days in which cable systems moved beyond 30 or so channels. It is infinitely more complex and important today with the essentially limitless amount of content, all of which can be viewed on multiple devices.
There were four winners in the category: Jinni, Tivo’s Digitalsmiths, ThinkAnalytics and Comcast. The technology is innovative and very important. Here are the highlights from the four products:
Jinni: Co-founder and CEO Yosi Glick was involved in entertainment software development for 15 years before gradually recognizing what many others in the industry did: The practicality of a grid consumer interface was fading as the numbers of choices proliferated. “We needed a whole new approach to content classification,” he said.
That insight eventually led to the Jinni system. It depends upon two technologies: machine learning and natural language processing. Jinni looks for 2,200 attributes in synopses of content. Out of those results, a taxonomy is built that describes the subject at hand. Glick says it is far different from simple search. For instance, the machine learning capabilities can decode the attributes astutely enough to steer viewers to a comedy about dysfunctional families as opposed to a drama about the same basic topic.
The national language capabilities enable the platform to understand what the person using the system is looking for no matter how the query is phrased. Glick said Jinni is on more than 50 million set-top boxes, including Comcast’s. Other users are AT&T, Vodafone and Telus in Canada.
Digitalsmiths: The company - which was bought early this year by Tivo - won its Emmy for its “Seamless Discovery” platform. Underlying the platform is a deep data repository that integrates descriptions of the content with behavioral data. Thus, a single record - which describes the data and how it is most often used - is used in an algorithm that determines the recommendations made to a particular user.
Similar requests coming from a tablet or a big screen TV may generate different recommendations, said CEO Ben Weinberger. The Digitalsmiths platform also is capable of delivering its recommendations to any device. “Some devices, for instance, might be more appropriate for searching, and others may work better with content pushed out” to the end user.
The platform is used by seven of the top 10 MSOs, Weinberger said. The five biggest users, he said, are Charter, DISH, Verizon, Time Warner Cable and Bright House Networks.
ThinkAnalytics: The key to the Think Analytics Recommendation Engine, said Chief Technology Advisor Christy Martin, is the ability to meet a subscriber’s desires across a number of devices. She said the platform is based on a predictive analytics and data mining.
Martin puts a lot of stock in the fact that ThinkAnalytics deeply understands how the cable industry works. Its policy engine - essentially, the quarterback - can be used to support the operator’s business goals. Suppose, for instance, that response to a query by a subscriber produces 10 possible recommendations. The policy engine can be programmed to weigh the recommendations presented toward those that carry advertising or in some other way support the operator’s business objectives.
Martin said the platform is used by Cox and Liberty Global, among others.
Comcast: The company's as-yet unnamed search product offers a number of features, said Rick Rioboli, the vice president of engineering for Comcast data services. These include ranking programs by personalized taste, full menu choices by taste, and generating recommendations based on favorite channels, shows and teams. "The system uses a variety of algorithms such as asset similarity ("these are both romantic comedies") and collaborative filtering ("people who watched this also watched that")," wrote Rioboli in response to emailed questions.
The platform is in wide use and is designed to evolve over time: "On next-generation cloud-based platforms like X1, we have the opportunity to regularly roll out new features and improved algorithms," he wrote. "We also have the ability to test and measure the quality of our results so we can continue to offer our customers better experiences, helping them find what they're looking for even more quickly and easily."
Making it easy for subscribers to vaguely describe what they want to watch - a movie “sort of like Breaking Bad” or a series that “has great political insights and snappy dialog like ‘House of Cards’ ” - is one of the most important tasks cable operators and their vendors face.
It’s an old problem, but one that is increasingly important and getting a renewed jolt of energy. “It’s sort of like SEO was a few years back, when everyone was focused on developing ways to find Web content,” Martin said. “We feel that as the UI becomes algorithmic that the way in which people find content becomes an important part of the experience - and of the business and marketing of cable.”
Carl Weinschenk is the Senior Editor of Broadband Technology Report. Contact him at [email protected].