August 1 2000 by Steve Bergsman
By many accounts, the next generation cell phone will have evolved, before long, into a kind of superdevice. Part phone, part PDA, part digital wallet?among other functions?the wireless devices of the near future promise to be the consumer’s primary Internet-connection-on-the-go. As such, their usefulness as mobile Web browsers will be determined by how quickly and efficiently they can scour the Net and retrieve information.
Anticipating that demand are companies like Autonomy that are developing the technology to empower mobile and other devices with just that kind of intelligence. The U.K.-based pattern recognition software company creates technology that extracts the meaning from any piece of text by recognizing concepts or patterns of concepts distributed throughout that text.
“This is jolly clever technology,” exclaims founder and CEO Michael Lynch, adding that it helps solve some of the limitation faced by today’s computers.
“At the moment all computers can do is spot individual words?a computer can’t operate on an idea level,” he explains. “But, what if the computer could read something and decide what to do next? What if human beings didn’t need to be involved? That is where the real value comes in. A lot of business can be automated if you can make the computer to be intelligent and understand what it’s looking at.”
What Autonomy’s software does is analyze a document, extract the ideas in the text, and determine which are the most important. In effect, it allows computers to automatically aggregate and organize internal (documents, e-mails) and external (news feeds) information into easy-to-navigate directories; categorize and tag all documents based on the ideas in the text; insert hypertext links to other relevant material; and access relevant data based on natural language queries.
“If you’re reading an e-mail,” suggests Lynch, “with the touch of one button, the computer reads the e-mail, understands and organizes it and related news or related e-mails without having to subscribe to all different directories.”
Obviously, text pattern recognition software can be adapted by end-user companies for a wealth of services, but in some regards it is also a labor-saving technology device. Since there is an information explosion of online content that Lynch says doubles every three months, companies can’t employ people fast enough to massage all that data. If text pattern recognition software can help the computer automate, less hands are needed.
Currently Autonomy’s technology can be found in almost all industrial sectors. Companies using Autonomy software include General Motors, Volkswagen, Pfizer, Morgan Stanley Dean Witter, and Merrill Lynch. Autonomy also sells its know-how to other software companies, including Sybase and Oracle, which build it into their own products.
Autonomy now boasts 250 customers across every sector. “We expect Autonomy to be in the center of many types of software for many different vendors,” says Lynch. “We are making good money in WAP in Europe and do a lot of work in digital TV as well as the Internet. If there is information in rich, human-friendly format, the computer needs to understand it. Digital TV or telephone?it does not matter to us; they are all big opportunities.”
In Europe, Autonomy is making inroads in the communications arena, with much business coming from the likes of News Corp., Associated Press, and British Telecom.
All this has helped the dually-based (San Francisco and Cambridge, U.K.) company to post earnings of $26.7 million last year. Sales growth was a striking 203.1 percent in 1999. Analysts expect earnings to climb to $60 million this year and there could even be some net income as well (the company did turn a profit in the first quarter). Last year, Autonomy lost $1.8 million.
“Our disadvantage is that this is a new field,” says Lynch. “Computers have not been able to do this up until now. People’s first reaction is to treat data manually and…this is all about automation.”
The big problem in e-commerce, Lynch continues, is that margins are getting very tight. “So to have 100 people sitting around typing or linking information is expensive. Now we are moving into the second phase of e-commerce automation.”
Text pattern recognition technology, says Lynch, converges into such new consumer fields as digital TV and older success stories, such as cell phones. In the former instance, a TV movie might show a beer can in a frame. Click on it and the consumer can buy the beer he’s looking at. “If you are on your mobile phone standing on a wet street corner in Paris and need information, and you type into some key word search engine from your phone,” you won’t get the information you need, says Lynch. “The phone technology needs to be more intelligent so that you can be reading an e-mail and with one button, bring up related news or closely related e-mails without you having to subscribe to all different directories.”
Lynch likens his particular field of search, analysis, and extraction to the aviation industry. The old keyword or manual operations were like propeller driven planes, while test pattern recognition technology is the advent of the jet engine. “While this is the first stage of jet engine to come out,” he says, “it is still way better than the propeller driven models that preceded it.”