Last year, HeyAnita was just another startup with a funny name that was developing technology to voice-enable the Web.Last year, HeyAnita was just another startup with a funny name that was developing technology to voice-enable the Web. Then, in April 2001, the Los Angeles company named after a friendly and knowledgeable waitress at CEO Sanjeev Kuwadekars favorite Hollywood diner landed a deal with Sprint PCS to provide voice- activated content and services to the wireless carrier. Two-year-old HeyAnita this month announced a third round of funding from Intel and other investors, bringing its total financing to date to $40 million. Matrix Editor Todd Spangler spoke with Kuwadekar. What have you learned working with Sprint PCS?
Sprint was the first major national carrier to launch these services nationwide. I cant share the numbers, but one valuable thing we got was some marketing data on what services people really like. Based on our experience not just with Sprint, but with another large telephone company were working with and from other data weve gathered the killer apps are personal information applications. We realized that early on. Applications like stock quotes and weather people use them once in a while, but on a nationwide basis, people want to check their e-mail every single day, two or three times per day. Another app they use is voice access to their address books, voice-activated dialing, voice access to their calendar information, things like that.
As a young company, how will you survive the tech industry slowdown?
We never really ran this company as a typical dot-com company. We were creating a software company. Eighty percent of our staff is in engineering. Because of that, weve kept our burn rates really, really low compared with other dot-com companies. What is important today is not advertising and marketing hype, but what kind of intellectual property you have and generating revenue. And that has been our focus all along. We have a lot of technology. Weve filed four patents. We have revenue coming in. We have an impressive list of customers, including Net2Phone, Sprint, Korea Telecom which is the largest telephone company in the country and Korean Air.
With all the revenue and with some of the deals were closing in the next few months, well be profitable in the first half of next year. How does speech recognition need to improve to gain wider acceptance?
There are really three things that need to happen before this technology becomes really popular. One is recognition accuracy itself. Companies like Nuance Communications and SpeechWorks [International] have done incredible work in this area. You can get accuracy of 97 [percent] or 98 percent, which is really acceptable as far as users are concerned. The second aspect is the voice user interface: It has to be extremely intuitive and very easy to use. Thats one of the strengths we have. One of the reasons Sprint selected us is because of the user-friendliness of our voice interface. The third thing is the killer apps. Companies that were competitors in this area launched portal-type services like weather and news. These are good services, but these are not the killer apps people will be using every day. Its being able to access your corporate data and being more productive. Thats what will drive adoption. So when will I be able to dictate an e-mail into the phone?
You can certainly compose a new e-mail on the phone right now, but today it goes as a .WAV file. Our system doesnt convert it to text. The accuracy of continuous speech recognition technology is not what were expecting. Its nowhere close to 90 percent. So until that happens, it wont make sense to do that, because it wont be very accurate. That definitely wont be by next year.