Content intelligence at scale. That’s the promise of the new Temnos platform from the company of the same name. Aimed equally at publishing networks that want to do a better job monetizing their output and brands with their own content initiatives, Temnos delivers metadata and metacontent for every URL sent its way, with the goal of helping the user understand its strengths and weaknesses, what audiences are primed to respond to it, how it might be repackaged to better direct to groups of readers or advertisers, what alternate headlines can be drawn out, and what summaries can be used, with the user’s choice of a more optimistic or pessimistic slant, low- or high-brow angle, and other features.
“It’s a good time to do this because both marketing advertising players and publishers and publishing networks are all hungry to make their products better,” says Temnos founder and executive chairman Tim Musgrove, who founded advanced semantic search and corpus analytics company TextDigger, which was acquired by Federated Media in 2010. Musgrove was Federated Media’s chief scientist in its Data Science Group until June, and Federated Media is one of the early adopters of the Temnos platform. “They find they can get a lot more leverage out of content marketing by doing this,” says Musgrove, helping to boost the CPM earned from marketers. “They can package their campaigns in a way that feels like it’s narrowing the targeting and not narrowing the inventory.”
Take targeting content for small business. Just plugging that in as a keyword would pick out a tiny fraction of the network’s inventory that it wouldn’t be worth an ad agency’s trouble to target against it. Keywords are no substitute for “true semantics and the construction of what a topic is all about and all the phrasing in which it can be couched,” he says. “With this we can quickly pick up 200 or more related topics to small businesses, including Quicklbooks, Schedule C, SBA loans – things that don’t fall right out of a taxonomy for free. We can grab hundreds of corresponding topics, know all these pages relate to small businesses, so Federated Media can offer a huge amount of inventory dead-on relevant to advertisers to target small business. That’s higher CPM but still with a lot of inventory. It’s a big game changer.”
As Musgrove sees it, Temnos 9.8 is delivering a comprehensive package to a space that so far only has had point solutions focused on solving a piece of the content intelligence issue to publisher networks and ad platforms, like classification. Kudos to companies like Peer39, he says, for selling those parties on the idea of using a classifier at all. But to date he says other vendors have had fixed taxonomies that almost never change; the Temnos service supports the IAB (Interactive Advertising Bureau) taxonomy, but it’s not limited to it, and its default reference technology Master Subject Index ties together the IAB and other taxonomies. Other efforts, he notes, have had simple Boolean true/false answers about whether a document belongs in a category rather than providing confidence levels of its relevance to it as Temnos does.
And also lacking has been extraction of Dublin Core Metadata. “We extract all that if not already in the CMS,” he says. “So we have competitors who’ve picked one thing to do in a static and fixed way, but we are more complete and flexible and customizable with some effort from development partners.”
The company’s heritage in natural language analysis and generation -- TextDigger drove CNET Review’s automated briefs for tens of thousands of products – play in its metacontent smarts now and to come. Among the capabilities of its algorithms are discovering more engaging headlines within the content itself and translating its grammar into headline news style, outperforming original headlines three-quarters of the time, according to Temnos. “We call it artificial intelligence but really it’s recognizing, reorganizing and repackage human intelligence.”
Musgrove also says that Temnos is one of the few companies with the technology to be able to write summaries of summaries for networks with tons of content – a Bloomberg.com, for instance, with multiple news sections and multiple stories per section each day. “If we can auto-generate product reviews for CNET about digital cameras we can summarize section by section news of the day for a major news outlet,” he says. “I’m pretty convinced that the natural language generation side is going to be big next year for us. We are looking at generating more of metacontent even then before.”
While publishers are a key audience, the trend to content marketing among brands means that they’re in line to consider such technology, too, to merchandise their output. Musgrove says Temnos is helping some major ones that create some very valuable content get more out of it. Take Tide as an example – they publish hundreds of well-researched pages about how to get stains out of every fabric. “Our algorithm might see an article by Tide ostensibly about how to get tomato sauce out of an apron but that this is the only way to do this that lacks toxins, and there’s a green message that’s not being exploited,” Musgrove says. “That’s a different topic, different possible headline, and maybe grab more people and more places and more eyeballs on to Tide than they realize.”