Radar Networks has been flying under the radar for a few years now, but on Friday the startup’s mastermind, Nova Spivack, unveils its first product, a brainy social network called Twine. Twine is a digital data dump with structure – what Spivack calls (what else?) a “semantic graph.” As with Facebook or MySpace, members get their own home pages, through which they can connect with friends or colleagues, and then they can create, on their own or together with other members, “twines” of information on any subject that interests them. A twine can include text, images, and videos that members either enter themselves or pull in, with a simple browser tool, from YouTube, Flickr, Amazon, or other online sources. Emails and other messages can also be thrown into the mix.

What makes Twine interesting is its Semantic Web underpinnings. (Spivack likes to throw around terms like “extensible ontologies.”) Rich with cross-referenced meta-data, the site’s underlying database allows Twine to interpret the information you put into it, automatically categorizing data, applying a set of relevant tags, following links to pull in related information, and so on. Its aim is to provide each member, or group of members, with a dedicated virtual librarian, working behind the scenes to organize a diverse or even chaotic array of information into a useful, structured form.

Twine’s database has been seeded with an initial dose of “intelligence,” mainly mined from the text, links, and tag clouds contained in various Web 2.0 sites, and its semantic understanding should grow rapidly as people use it and add more information and tags. Like Google’s search engine, Twine provides a handy service to its users and then systematically harvests intelligence from those users to continuously improve the service.

Business model? It’s a work in progress. In the near term, expect the obligatory AdSense ads. Longer term, Spivack is counting on using the site’s insight into its users and their interests to develop a custom system for serving highly personalized ads and product recommendations. More interesting, Twine will offer a for-fee professional version (the basic version is free) aimed at business people and, in particular, teams of workers doing research or otherwise tracking and analyzing information. Spivack says he plans to allow outside services to tap into the Twine database through open APIs, although heavy users will likely have to pay for the privilege of incorporating the Twine brain into their own products.

Spivack says that Twine is not intended to compete with Facebook and other social networks. But while it’s true that Twine is a different sort of thing, it’s also true that it promises some compelling information-management benefits for business users that Facebook can’t match. If one of the assumptions behind Facebook’s rich current valuation is that it will become a popular business platform for sharing ideas and information, then Twine poses a clear and imminent threat.

I’ve only seen a demo of the site – a restricted beta begins Friday. The demo was impressive, but that’s to be expected. What remains to be seen is just how smart Twine really is. If there’s a lot of noise in the information and tags the site generates, it could frustrate users. There’s nothing more irritating than a dopey artificial intelligence. But, with Twine, Radar Networks does at least seem to have given itself a leg up on Metaweb’s Freebase, the other high-profile attempt to build a mammoth semantic database to power the Web 3.0 world. To use Freebase, in its current form, is to play the role of an unpaid data-entry clerk. It’s boring. Twine at least gives you a reward as it picks your brain.