Jennifer Zaino
SemanticWeb.com Contributor

Semantic intelligence and predictive analytics startup BeliefNetworks this week announced both a partnership with online publisher Managing Automation and the launch of its Affinity Agent Recommendation Platform for enhancing advertising and content placement.

Managing Automation says it plans to use the platform to improve the relevancy and locations of ad placements on its site, and generate greater interest from users in interacting with advertisers, according to a statement. The platform, according to the vendor, provides a patent-pending concept mapping and learning technology that helps users find semantically related information, such as news, videos, music and images, while enabling Web site publishers and bloggers to access and link to related content of their published subjects and areas of interest.

"Managing Automation didn't want to have to think about which ads were optimized for their portal," says BeliefNetworks CTO Ted Tanner Jr. "And they didn't want to belabor the people advertising to come up with proper optimized keywords, so we automate that process and optimize based on the content of the page."

The news follows the announcement last month that the startup had so far raised $1.4 million in angel investments in a still-open seed fund round. Target markets for the company's technology, built on its DataChasm platform, are online advertising content and related personalized content, as well as public safety and homeland security applications.

What's the link relating its technology to these varied sectors? "We look at everything as finding anomalies," says Tanner. Anomaly detection connects purpose, whether it is increasing the click-through ad rate for a page, tracking an ensemble of concepts across Twitter to better market a product, or finding that bad actor in the public safety space in a sea of bad actors, he says.
Additionally, "advertising networks are large arbitrage environments where you put up as many ads as possible as fast as possible. Public safety and homeland security environments are sometimes the same -- they are completely real time and [about] finding someone immediately based on profiles or actions or concept linkages if you will." So, one of the tenets of the platform was to make it run as fast as possible.

"That's a hard goal for many learning systems because historically you had to model these things overnight." Tanner contends that a smaller computational footprint to do categorization and predictive analytics, speed to deployment and accuracy distinguish its platform.

The platform, which runs in a cloud computing environment, is distributed on the Hadoop open source implementation. Recently, BeliefNetworks also has taken its technology to the consumer arena, in an alpha application dubbed MashMeUp. Enter a phrase or web site at Mashmeup.com and it returns concepts related to that query. You'll get results of who's talking about those concepts at Twitter, web pages and FlickR images related to those concepts, and combinations of these.

"Our main business now is B-to-B, but we had lots of requests to make a consumer portal," says Tanner.

Still, BeliefNetworks is taking care about what spaces it will venture into. Tanner, a veteran of numerous startups, says the company has had interest from health organizations and hedge funds about applying its technology to their requirements, but he says it's important for a startup to stay focused on its core missions.

"One of the things I've seen many startups do is become defocused, and take every customer that comes in the door," he says. "You have to be focused, you have to scale according to need. Because we are building a platform, we see a future in applications in several areas, but we want to be very precise in showing what BeliefNetworks can accomplish and making it real."