Rovi is marrying man and machine.

Taking a page from the big data playbook, Rovi Corp. has introduced a standalone product called the Rovi Knowledge Graph that combines its traditional curated data sets with dynamic, machine-generated information. On the curated side, Rovi still uses 300 human editors worldwide to build its metadata core. On the machine side, the company has added in data from 100,000 online sources that dynamically track news and entertainment updates. Those sources give Rovi customers "access to 20 million additional keyword assignments, 545,000 additional sub-genre assignments, 920,000 additional assigned roles and metadata on four million additional people."

The knowledge graph approach to TV metadata is significant for two reasons. First, it ensures the most relevant and up-to-date information is available for content search and discovery: Information such as which new TV clips are popular, or how different Hollywood stars are connected to each other. Searchable information isn't limited to just 14 days of guide data.

Second, the knowledge graph makes it possible to determine links between different data points -- for example, when an actor is closely associated with a specific genre of content, or when a director works with one or two of the same actors repeatedly. This relational data helps with content recommendations, and with the ability to provide a conversational experience to TV viewers.

Want to know more about the impact of web services on the pay-TV sector? Check out our dedicated OTT services content channel here on Light Reading.

Although Rovi's new metadata product is a standalone offering, the knowledge graph technology is also integrated elsewhere with the company's Conversation Services solution. By putting the two together, Rovi can power an interactive experience where users literally talk to their televisions and get intelligent responses in return. Senior Director of Product Marketing Charles Dawes explained in an interview that Rovi's technology even makes it possible for a user interface to retain the context of a conversation over the course of several sentences. If a user asks a question about a movie and follows up with a second question using only the pronoun "it", a Rovi-powered UI can recognize the reference and answer accordingly. (See Rovi Makes It Personal.)

If the conversational experience sounds familiar, it's because Comcast Corp. (Nasdaq: CMCSA, CMCSK) is highlighting a similar technology with its Xfinity remote with voice control. At last week's Internet & Television Expo, Comcast CEO Brian Roberts showed a demo of the talking remote that included picking out an individual scene from a movie, and searching for biographical information on a related actor. Roberts also announced that the talking remote is now available to all subscribers across Comcast's footprint.

Comcast is notoriously mum on many of its vendor relationships, but Light Reading learned in 2013 that the company was using Jinni Media Ltd. 's technology for content recommendations in its X1 platform. Comcast has also traditionally licensed metadata from Rovi. (See Jinni Powers Comcast X2 Recommendations.)

— Mari Silbey, Senior Editor, Cable/Video, Light Reading