PureDiscovery wants to remake search in the brain’s image, raises $10M to do it

PureDiscovery, a Dallas-based startup trying to reinvent enterprise search via its BrainSpace technology, has raised $10 million series C round led by Medina Capital. Rather than indexing documents and letting users perform keyword searches, PureDiscovery is focused on semantic technology and learning the concepts contained within a company’s content.

We first covered PureDiscovery in early 2012, when the company was just beginning its push out of the legal field where it has already made a name for itself. The company’s software has proven effective in e-discovery, where it’s used to learn what’s contained within thousands of pages of documents turned over during litigation. PureDiscovery also powers patent search for LexisNexis, where it has analyzed hundreds of millions of patent documents and journal articles to surface the most-relevant content regardless of keyword relevancy.

An initial scan of documents in e-discovery.

An initial scan of documents in e-discovery.

With the latest investment now a done deal, PureDiscovery Founder and CEO Dave Copps told me it’s ready to make that push into the broader enterprise market. In order to do that, it’s moving away from traditional software and into a cloud-based platform, and is about to release its general-purpose BrainSpace technology (previously called Grokkit) to the world.

The idea, Copps said, is to create interest graphs linking people and documents with concepts. So if I’m interested in “big data,” for example, BrainSpace should point me to internal company documents about that topic, but also to external content and to people who know a lot about it. Ideally, it identifies people based not just on keywords in their profiles, but on their interaction with big data content and perhaps their activity on the corporate social network.

PureDiscovery's current product, clustering documents by their content.

PureDiscovery’s current product, clustering documents by their content.

“It’s not about who knows who in the enterprise — that works fine for Facebook(s fb_ — it’s who knows what,” Copps said. He added, “It’s not a search problem, it’s a connectedness problem.”

Copps gave me a demo of the general-purpose BrainSpace at our Structure: Data conference in March, and it was a fairly impressive product. If it works as demonstrated (and if my memory suits me), after finding one piece of content, he highlighted a passage from it, dragged it into “the brain” on the screen, and it brought up a whole new set of content relevant to that idea. Theoretically, Copps explained during our first discussion last year, this same technique could work for consumers across the broader web, too.

Such a broad platform would have to look and function a lot more like Google than like enterprise software, but it would help usher in the next generation of search where content finds users instead of the inverse. It’s like Google’s Knowledge Graph features taken to the next level (although delivering anything at Google’s scale is another challenge altogether).

The product is supposed to be available as a beta release in several weeks, Copps said, so we’ll be able to see then whether it’s living up to its promise.