Sramana Mitra: Can you succinctly explain what it is that you have that these large companies don’t have that enables you to win those accounts?
Dave Copps: I’d say it’s two things. It’s our core technology and our user experience. We have built a technology that can learn dynamically without the use of any human intervention. Most companies out there, even the big ones with millions of dollars in funding, have some sort of a human-built portion – a lexicon or ontology – that backs their technology. We are one of the first companies to come out with this idea of completely unassisted learning.
You give me a million documents from your company, I’ll read those million documents and build them into an intelligent brain that everybody can use. We built a core technology that can learn dynamically from unstructured information without any human intervention, and that makes us very unique. One of the advantages it gives our customers is they can walk into virtually any company without any pre-work. We can be up and running in, literally, hours where other companies take weeks and months to get their technology up and going. To give you an idea, someone who has a million documents and wants to build a brain with Brainspace, that will take about 40 minutes to build.
Sramana Mitra: That’s very valuable. I know this kind of technology reasonably well. I can appreciate how complicated this is and how much value you’re providing in the customer context. That actually answers my question. That is a significant leg up. Would you be comfortable discussing a bit more about the technology itself? What is in the computer science that enables you to achieve that?
Dave Copps: Semantic technologies, typically, use a document-to-document comparison. You find a document and you say, “Find more like this.” Basically, what that technology does is it looks for a semantic centre. It averages the concepts in the documents and looks for another document that has the same centre. That’s really a flaw in semantics. What we found out is, obviously, people don’t work that way. If you read a document, you don’t average the concepts. You actually remember two or three things about it.
We basically, at a very fundamental level, did something different. We don’t do a document-to-document comparison. At a very fundamental level, we do a document-to-word comparison. What we can do that other people can’t do very well is, what we call, transparent concept search. You give my system a word for phrase and we’ll actually give back to you all the words and phrases that we think are related to that even if they weren’t in your query.
Sramana Mitra: Where is that coming from?
Dave Copps: Essentially, it’s a hyperspace. If you give me the word Java, I’ll find the word Java in my hyperspace and I look at all the words and concepts that are harbouring around it.
Sramana Mitra: The hyperspace is the Internet?
Dave Copps: No, it’s ours. It’s a multi-dimensional index that we built from documents. When you give me a million documents, my build process is that I run it through a set of algorithms that basically transforms the documents into a hyperspace of between 275 and 350 dimensions. It’s literally like stepping into a sphere and looking around you and seeing words and phrases.
Sramana Mitra: Are you creating this hyperspace for each client?
Dave Copps: Yes.
Sramana Mitra: Is it an instance of hyperspace per client?
Dave Copps: It can be, yes. We have both. We have built a brain on the web and here’s what we did. One of my early visions was, “What if we could build a brain on the web?” That’s something that’s really driven this company. We came to the realisation that the idea needs a little changing. Brain in the web is a good idea but most of the documents on the web are complete crap. 80% of everything out there is not that valuable. How do you find the best documents on the web, and could you build a brain from that? We, basically, have done that. We decided to rely on social media.
Our theory is that the best documents on the web are the ones that are being shared on social media – the ones that are being retweeted the most, the ones that are voted the most, and things like that. We formed a relationship with Twitter and other social services. We’re bringing down about 350,000 new documents a week right now. Essentially, we’re putting them into a bucket. That bucket is over 40 million documents right now and it’s growing at 350,000 a week, and we build a brain on it. We don’t choose the documents. We let the social graph choose the documents.
We connect the concepts and all the documents into an intelligence of Brainspace. Then, everybody gets to use it. Next time you’re looking at a document, you can highlight a paragraph, put in a keyword and Brainspace will show you other words that are related. You get to interact with the brain basically. It’s almost like augmented intelligence. It’s a little bit different approach. Most semantic search engines are blackboxes. They bring you result but they don’t tell you how it got there. We’re exactly the opposite. We expose what the brain has learned about and show you the related concepts, and let you move slider bars under each term to say more of this and less of that.