Sign up for our newsletter - Navigating the horizon of business technology​
Technology / AI and automation

DICTATE INTO YOUR COMPUTER AND TALK TO YOUR TV

If we are to believe industry predictions, speech recognition technology is going to be big business in the not too distant future and recent months have seen numerous organizations pledging their commitment to this growing trend. IBM Corp has recently updated its offering with the introduction of ViaVoice (CI No 3,180), while software giant Microsoft Corp has just announced a $45m investment in Belgium speech recognition outfit Lernout & Hauspie Speech Products NV (CI No 3,244). And Dragon Systems Inc, which has just unveiled a UK version of its NaturallySpeaking speech recognition dictation software (CI No 3,241), believes we are going to see an explosion in the voice recognition market similar to the speed at which the internet has grown, as the technology spreads to the personal computer. The Dragon application has been designed to enable its user to talk into a microphone and see their spoken words translated into text, appearing on the screen of their PC. The company has been specializing in speech recognition technology since its inception some 15 years ago. Dragon claims NaturallySpeaking has accuracy levels of at least 95%. In order to run the application on a personal computer the user must have at least a 133MHz Pentium machine with 32Mb of memory, running Windows 95 or NT 4.0, but the company recommends a 166 MHz Pentium with 48Mb of memory, purely because it will perform much faster.

By Emma Nash

The Boston, Massachusetts-based company describes the product as a general purpose large vocabulary continuous speech recognition system for PCs. It has a vocabulary of some 260,000 words and users should be able to customize the product to their own voice within 18 minutes. Dragon has 160 researchers and developers who have integrated a cross section of UK accents into the software which should ensure effective use by 99% of the population. People with particularly strong accents, or those who have difficulty with the product churning out a higher than expected rate of errors, are recommended by Dragon to spend a little longer training the product so it can get used to the users voice. The company launched the US version of the product at the Comdex show in New York in June and since then has had to speed up its production process so as to be able to cope with demand, with some 50,000 copies being sold in the first two days of its availability. IBM recently updated its speech recognition offering with ViaVoice which has a retail price of around $200 in the US (CI No 3,180), and is being launched in the UK this week with a retail price of 100 pounds. Dragon obviously believes it has a superior product to that offered by IBM, and is not at all put off by the size of its nearest competitor. Andreas Widmer, Dragon’s director of international sales and marketing believes it is an advantage to have such a large competitor with a big advertising budget because it raises the profile of the market as a whole, and proves that it is not just one company on a one product mission. But IBM’s UK speech sales and marketing executive, Elaine Richards, unsurprisingly believes IBM’s product is superior and will be more attractive to the market because it is considerably cheaper. Richards says ViaVoice, like NaturallySpeaking has an accuracy level of 95%, but can only sensibly deal with text delivered at up to 140 words per minute. But she points out that ViaVoice can be delivered straight into word documents, meaning users don’t have to move their completed documents into standard word processing packages. ViaVoice is aimed at the corporate market, and Richards believes products such as these will not be successful in the UK market if they are priced any higher than the magic 100 pound mark. IBM’s consumer offering, SimplySpeaking has been in the shops for some time, capturing 80% of the market share, and retails for half the price of the corporate model at 50 pounds. Richards said IBM’s aim is to make its product available to as many people as possible. With a product selling for more than twice the price of its larger rival, Dragon appears to face a tough task. But Chris Stimpson, the company’s international marketing communications manager said it is an apple and orange situation because the IBM product has yet to come out with a UK English version. He also said that judging from the response the product has received in the US, Dragon has no plans to reduce the cost of its product. Dragon’s NaturallySpeaking product is the ‘Personal’ version which enables the application to be used by just one person, but the company is currently working on the Deluxe version which will be out later this year, enabling unlimited user access. This particular product currently only runs on Windows, but Widmer says Dragon is talking to Apple Computer Inc at the moment about a possible Macintosh version in the future. The emphasis is on saving time, and NaturallySpeaking can apparently deal with speech delivered at up to 160 words per minute, and although there is a potential market consisting of anyone who has a PC, the product has proved popular among the medical, legal, government and general business sectors, with a high take up also coming from students and people running small or home offices.

Explosive growth

White papers from our partners

Both Dragon and IBM foresee huge growth in the speech recognition market and believe it is about to witness explosive growth. The entrance of Microsoft into the sector is also likely to fuel attention and awareness. Richards says it is important to recognize that the speech recognition sector is already a big market, but she believes it has benefited immensely from the transition from discreet speech technology, which requires a 10 second pause between each word, to the phenomena of continuous speech that the industry is seeing today. Who manages to out-sell whom, seems somewhat irrelevant at the end of the day. Both IBM and Dragon believe speech recognition will broaden out from just word processing, and it won’t be long before we are talking to our television sets. What is certain, is that in the not too distant future there is more than a distinct possibility that we won’t have to sit in front of our keyboards and tap fanatically. We can sit back and simply talk to our machines.


This article is from the CBROnline archive: some formatting and images may not be present.

CBR Staff Writer

CBR Online legacy content.