Seen anything good on TV lately? If the answer is ‘No,’ then maybe the problem is that you and your TV just aren’t communicating as well as you could be. The same may be said of your experience across other viewing mediums, like smartphones, tablets and PCs.

Veveo wants to change the picture, so to speak. “We want the TV to be as friendly as possible so you and the TV can have a really productive relationship,” says CMO Sam Vasisht. The company, which earlier this month exhibited its Conversational Interface Technology at TV Connect 2013 in London, says there’s a need for a universal interface based on natural language capability, so that people more intuitively can grasp what is available from where in a world of fragmented content sources, including how to better search for that content and manage their viewing experiences with greater speed and ease.

“Voice is probably the most natural way for us to deliver this experience,” says Vasisht. Veveo wants to be the platform that enables service providers and OEMs and video programmers to give their audiences the power of speech. User interface designs and delivery mechanisms (i.e. built into a set-top box or delivered via a mobile phone) and speech-to-text engines are left up to the customer, while Veveo provides the natural language processing technology and the ability to learn users’ behavior to better predict what the user is looking for in his search. A semantic knowledge graph is a key part of it, too: “You know the semantic connotations [of a search] over the Knowledge Graph,” says Vasisht.

The company’s take on NLP is that it works best when focused on a specific vertical app – and it’s going with video for starters. In comparison to a more general approach like a Siri takes, “we can go very deep to get a very intelligent understanding of what a human says no matter how they say it,” he says. That way, Vasisht says, a query about “who are we playing tonight” is understood to be about a game being broadcast on TV, not a callout for what movies are playing in nearby theatres or which version of Angry Birds you want to play.

At the same time, it knows enough to disambiguate between like terms. If you ask to see the Sox schedule for the evening, for example, it can question whether you want the White Sox or the Red Sox if more than one game is going on. It also applies its growing understanding of user behavior to inform search responses, so that it ultimately can make the decision that you mean the Red Sox even when both teams are playing, based on past actions. (Multiple users in a household can be disambiguated by provider-deployed features, like colored buttons on remotes assigned to individual family members, or by time-of-day and day-of-week – very early weekday mornings, for instance, likely belong to parents and news shows, while kids have cartoons in the after-school hours.)

The context of conversations is maintained, such that once it asks you which Sox team you mean and you answer Red, it immediately gives you the answer about when that team is playing, rather than requiring you to restate a query. Or, because the system maintains the context of conversations, you can go even further, says Vasisht. You can say, for instance, show me action movies, and take that through James Bond movies only, to James Bond movies with Sean Connery, to other Sean Connery movies beyond the James Bond series. “You can drill down and up the context based on a continuing dialogue with the user,” he says.

Knowledge Graph At The Root

That brings us to the semantic Knowledge Graph. “Our cornerstone of what we do is the Knowledge Graph that aids with semantic interpretation,” says Vasisht. “We can predict based on how close the links are in the Knowledge Graph, how close queries are to what has already been discussed.”  Veveo began building its Knowledge Graph nearly a decade ago, and it now accounts for more than 100 million nodes.

“We have applied the Knowledge Graph, which is the common denominator for all we have done, to 100 million devices today,” he says. The non-speech enabled version of its search and recommendation technology for content is used, for instance, by some 45 million TV sets connected to services from Comcast, Cablevision, DirecTV, FIOS, and the like. Nokia also leverages the universal search on some 35 million mobile phones.

The Knowledge Graph is updated in real time on an ongoing basis. Providers determine how they’ll leverage that, though. For example, cable operator customers tend to run it in theiw own networks vs. Veveo’s cloud model because they want full control and privacy. In that case, Veveo provides updates based on whatever the SLA is, which could be daily, weekly, or so on. “But as we get into more dynamic aspects of TV programming, we can provide it on an hourly basis if they want – for example, if with Twitter and FaceBook feeds you want to capture trends very quickly,” says Vasisht.

Europe is fertile ground now for the company, he says, as U.S. operators have take the lead as far as rich TV usability goes. Also, it’s worth noting that the company isn’t ignoring the direct consumer market  and accessing over-the-top content delivery, but there’s nothing new to formally say at this time. (Today it offers consumers apps like its vtap QuickSearch for Android universal search for finding — not via voice — apps, music, contacts and such). Meanwhile, it continues to accumulate patents, adding three more in March that gave it a total count of 34 patents issued by the United States Patent and Trademark Office.