Say It on the Internet

Friday, Feb 1, 2002
Corporate websites are not places that Net surfers tend to visit for fun. But Rhetorical Systems' site (www.rhetoricalsystems.com) could prove an exception. The Edinburgh-based text-to-voice software company has a demonstration page that lets users type in a short message and pick one of several different synthesized voices — from a Scotsman's burr to a Californian Valley Girl's twang — to read it aloud. The results can be amazing and amusing. The female, southern English voice was, for instance, spot-on in pronouncing the Mary Poppins' song title, Supercalifragilisticexpialidocious, if not quite in Julie Andrews territory.

Marc Moens, CEO and a computational language engineer, co-founded the company in 2000 as a spinoff from work he and colleagues were doing at the University of Edinburgh. They launched their first product, rVoice, in May of last year. The latest (and third) version of rVoice offers a dozen natural-sounding voices in several dialects. But the software can also be customized. So a client could request and get, for instance, a voice that spoke in an American Southern drawl or an Irish brogue. Moreover, a specific individual's voice can be synthesized, as well, once enough of it is first captured on tape. Moens says that Rhetorical Systems' ability to create dialects and do them quickly is what gives it an edge over competitors.

There are myriad application possibilities, particularly accessing Internet data from a phone, especially a mobile phone. The software can give voice to e-mail and faxes, in-car navigation systems, and online games. Electronic tutors at e-education sites could give students vocal feedback, and animated talking heads could let users hear as well as see information online, perhaps letting an e-tailer verbally point out sale items while a customer is browsing. And, of course, there is an obvious market for helping the visually-impaired gain access to the Internet.

Moens says Rhetorical Systems not only sells the software, but advises clients on how best to incorporate it. The e-mail application can, for example, also include technology that lets users speak commands to skip paragraphs, repeat segments, slow down its cadence, and prioritize their e-mail. The algorithms used in rVoice will correctly repeat most common-usage words. If it comes across an unusual word or a nonsense word, it will make an intelligent stab at the correct pronunciation, and often get it right.

Rhetorical Systems in November got a fresh infusion of cash, a private placement of $6.2 million that Moens says should see it through to profitability in about a year's time. It already has offices in New York and Boston, and has begun marketing in Australia, as well. And before mid-year, it expects to roll out its first foreign-language version, though Moens isn't ready to say which language it will be. The global market for synthesized voices is forecast to reach $6 billion within a few years. Talk may be cheap, but for Rhetorical Systems, giving voice to the Internet should prove financially rewarding.