Tuesday, March 20, 2007
Interview with Scott Jarus, CEO of Cognition
Scott Jarus is the CEO of Santa Monica-based Cognition (www.cognitionsearch.com), a startup in the search engine space. Jarus is the former CEO of J2 Global and an angel investor, which is how he ended up at Cognition. The firm is announcing the commercial release of the firm's technology today, has rolled out a publicly available web search product at its web site, and is offering its technology to enterprise customers in the litigation support and life sciences markets. We sat down with Scott yesterday to understand how the firm fits in the search space, how it is different from the search engine technology currently available, and how it hopes to use its technology to help users. Ben Kuo spoke with Scott.
Thanks for the interview. What is Cognition?
Scott Jarus: We have developed the next evolution in search engine technology. What we have created, is a meaning based linguistic search technology. The company was founded by a team of Ph.D. rocket scientists--our founder, for example, is a Ph.D. in computation linguistics and a post doctorate in Computer Science, and a noted author at UCLA, Kathleen Dahlgren. Our technology has been in development for about 20 years. In the early years, up until about four or five years ago, the technology was a research and development skunk works project. It was a development project in linguistic search, which had been sold as components to companies like Motorola, the Department of Defense on contracts. Four or five years ago, the technology reached a new level of development, about the same time that Moore's Law for computation really kicked into full gear, and Moore's Law for storage kicked in. It was a perfect storm - Moore's Law for Storage, Moore's Law for computation, and the technology came together at the same time. We then worked on the product, sizing it and making it available as a commercial product.
Can you talk a little bit about your background and how you ended up at Cognition?
Scott Jarus: I joined the company about a year ago. I'm an angel investor, and found the technology and business opportunity so compelling that I not only invested a significant amount of money, but also took over and have been running it as CEO since. The background is, I was the CEO of J2 Global, which most people don't know by that name but for its consumer service, eFax, which is the FAX to email service. J2 Global is a public company. When I joined in 2001, it was a badly broken company with a market capitalization of $27 million, of which $14 million was in cash--which means the enterprise value was only $13 million. Four and a half years later when I left the company, it was worth over $1 billion in market value, and the stock price went from $2 to $42 dollars, and we had $110M in cash at the time. We went from 4 million to over 9.5 million customers. It was a very successful turnaround, which gave me the luxury to become an angel investor and find Cognition Search. Other members of the team include a VP of Sales and Business Development who has experience at Dialog, the large content provider and search provider in the enterprise space, Knight Ridder, and Infotrieve--Infotrieve being a noted provider of biomedical content for healthcare and life science industries. The company was funded by angels, and has several million in angel funds--coming both from myself, a smaller amount from the founder, and a significant amount from the Tech Coast Angels, which is a organization of angel investors here in Southern California. I am also now in the Tech Coast Angels, though my first investment was not as a Tech Coast Angel.
I mentioned that this is based on meaning-based, linguistic search technology. In essence, we have taught the software the meaning of words in human language. It understands what a query is and what is contained in documents being searched. What differentiates this from most other search engines, is they don't understand meanings of words. What it is commonly compared to is a Google or a Yahoo, which are pattern matching. We have built this over 20 years, with 300 plus man years of development work. It took so long, because we had to hand code each word and concept in the human language to make it work. That was very laborious, but it is done now, with 99.9% of common English language done. Of course, we're never completely finished, because there are two hundred or three hundred new words added a year. The technology is ours, we haven't licensed it, and we completely developed it ourselves. It's also patented. The technology increases the precision and recall of search results, addressing simultaneously data overload and underload. We've attempted to mimic the human language as close as possible, using various disciplines of linguistic science, and we believe it's the most advanced search language today.
How does the company fit into the search market then--do you supply your technology to others?
Scott Jarus: We currently market this product in two different ways. We offer it in the vertical search channel, and in the legal and life sciences areas. In the legal area, it's in the submarket called litigation support. There's a software package called Lexis Nexis Concordance, whose largest use is for the legal market, with over 65,000 desktops. Lexis Nexis came to us looking for advanced search capability. They currently only have a pattern matching search engine, and we partnered with Everquest to develop a product which provide advanced search capabilities using Cognition's meaning based search technology. We've also focused on the life sciences industry, where they have the problem of dealing with vast amounts of data, most of which is proprietary and internal. Those are the ones that the customers who are using us now. Last Tuesday, we launched a web search portal, which is a technology demonstration focused in two or three areas--legal, politics, and health, where you can do a very deep search in those areas.
So are we going to be seeing your technology mostly in these verticals, or will users see a general search engine with your products?
Scott Jarus: Right now users can use our specialized web search portal, which is a demonstration site. It will expand, until we have sufficient funding in the company to index the entire web. Web search engines are very expensive, and we don't have the funds to do that. We're currently offering very deep content in health and politics, and other content categories which we will be filling out in the future. The content is real, and we will be adding more content. The demonstration site is there to show value, and our technology. As we gain users and critical mass and eyeballs, we'll be able to monetize the site with advertising and things like that.
The site is really complementary to existing search engines. Pattern matching search engines have a very good place in the marketplace. Our technology is augmentative to pattern matching search, allowing you to dig deeper into particular topics. If you are searching on Google for a hardware store, you'll get those results right away. However, if you search on Google Type 2 Diabetes for men, you'll get 16 million articles and overload. There should instead be a way for you to specifically dig down for Type 2 Diabetes for men. We can provide higher recall as well.
How is your technology different from the techniques of automated term expansion and substitution that are currently being used by pattern matching search engines?
Scott Jarus: We have technology well beyond that. The problem is beyond automated text expansion - which is called thesaural enhancement, using words of similar meaning. The problem is, synonyms have synonyms themselves. In the case of complex statements--say, you are searching for "energy initiative" it may not have known that "energy initiative" is a piece of energy legislation, a program to reduce energy. The power of our technology is not only the understanding of the individual words, but words within context of a query and document set.
There are a lot of search firms and grad students projects that are taking on some of these issues, why is Cognition different?
Scott Jarus: First of all, we've got the soup to nuts--a parser, search engine, and indexer--and while some companies may have a piece of it, we have a 20 year head start on linguistics, meaning based search. Others don't have the technology, or ability to scale. We think we're miles ahead of the nearest competitive technology. We've already launched, and we know that to develop the parser, index, or dictionary certainly would take an amount of years--not 20, perhaps, but certainly not one or two, more likely five to seven years to develop the language capability.
Our differentiation from everyone else is also that we have taken the academic science, and R&D, and commercialized it. We have a product which is now being sold and used both in vertical enterprise and the web. The biggest differentiation is that the technical side of the business--which was twenty years in the making--needed to be augmented by people who understand the business of search. That's where I come in, and our VP of Sales and Business Development, and several others on our team.
When, if ever would we see your technology in a general purpose search engine like Google?
Scott Jarus: We're an augmentation to what people are typically used to. We're not representing that for a large percentage of searches on the web, people would have a desire to use a linguistic meaning based search engine. Most searches on the web are 2.1 words. They're looking for real quick information, the first five results of a pattern matching engine. However, we provide search for those individuals who are doing research on something of importance to them--ranging from not only health care, but information about a particular car or problems with a particular car, or a political blog and searching beyond pattern matching. The power behind our search engine is additive and augmentative to pattern matching search engine, and allows you to dig deeper--search less, and find more. That's really the nut of it here. You don't have to go through hours and hours of digging to find what you're looking for. We can do it faster, and with less energy spent looking through content.