Archive for July, 2005

Article in Flemish newspaper De Tijd (July 25) on speech technology

Monday, July 25th, 2005

The “Wetenschap & Technologie” section of today’s newspaper De Tijd features an article by William Visterin entitled “Spraaktechnologie, vijf jaar na de val” (= “Speech technology, five years after the fall [of L&H]“). I have had the pleasure of contributing to this article as one of the two interviewees. The other one is Marc Moens, founder of text-to-speech technology provider Rhetorical Systems (now a part of Scansoft).

As far as I know, the article has not been published online for non-subscribers. Here’s my own summary:

1) Dictation software remains a niche product, used e.g. for automation in the medical and legal domain. In the consumer market, there is less interest: unless a PC user has a compelling reason like a physical disability, competition from mouse, keyboard [and screen] remains too strong for mass adoption of speech technology as an interface to a PC.

2) Embedded speech technology (in mobile devices or in-car navigation systems) is mentioned, but without any further elaboration.

3) On the advancement of the basic technology over the past five years, increased availability of processing power has allowed for the retrieval of sound segments from larger and more refined databases, resulting in higher-quality text-to-speech systems. In speech recognition, processing power has contributed to better accuracy. In multilingual markets like Belgium, the availability of simultaneous multilingual speech recognition engines is an important breakthrough.

4) Half of the article is about the lucrative application of speech technology for automating routine tasks in a call center environment. The author mentions the tribulations of Belgacom’s virtual operator Lisa (”1234″), and the lesson that the speech technology field has drawn from this and other live phone applications: the increased focus on dialog design, human factors, man-machine interaction, in short, on usability. Achieving user-friendly systems does not only require better development tools – available on the market now – but also specialised knowledge, which is in shorter supply.

5) On the acceptance of speech technology in the phone network, the author quotes Marc Moens, who states that “Ten years ago, many people disliked automated answering machines. Today, we all speak in our messages in each other’s voice mail”.

All in all, the article gives a succinct but objective overview of the state of speech technology and its applications in Belgium anno 2005. No more hype, but no gloomy messages either, just reality. That in itself is a big step in the right direction.