PCs Go Chatty

It can type faster than any typist you've ever known. It can com-pose email messages, draft letters, create reports and access information from your database-without even using a keyboard or a mouse. All you have to do is just speak to your computer, naturally.

With the way the computers are penetrating into our day-to-day lives, it is clear that a faster and more effective mode of communicating with them is needed. Speech recognition (SR) and Natural Language Processing (NLP) systems are becoming increasingly popular across the world. "There are many people who want to use speech as an interface with the computer. Instead of punching on the keyboard, if SR is used as a tool to navigate, it will make things much easier," says Vanit Arora, Country Marketing Manager (Software Solutions), Tata IBM.

Sound success





One major reason for this optimism about SR software is that technology has moved toward recognizing continuous speech. Earlier, every word had to be spoken separately with long pauses. "From the vendor's viewpoint, it is now the right time to push this technology. Since IBM has a wide range of software, we work on the basis of a product mix, which needs to be focused. Till last year, SR software was not one of our priority areas," says Arora.

"Numbers seem to justify this new-found confidence among the vendors. Dragon Systems, a worldwide leader, which entered the Indian market only a few months ago, sold about 800 units in two months," says K Shivakumar, Principal Operating Officer, HCL Infosystems. The company which has a distribution tie-up with HCL hopes to sell 4,000-5,000 units in the next year. According to Arora, IBM's voice-type dictation product recorded sales of about 2,500 units last year. The company expects its sales figures to grow by 200% next year.

Dragon, which enjoys a 50% marketshare in the US, has shown promising sales of 1,41,000 in the first half of 1998. IBM sold 41,000 units for the same year. "In the US, it took four-five years for the sales to pick up. But in India, we expect it to happen much faster because the product has now become extremely user friendly," said Graeme Vanderstoel, Director, Asia-Pacific, Dragon Systems.

The company's 'Naturally Speaking,' recently came up with its version 3.0. The product is available in three models-standard, preferred and professional-depending on the number of features available. Dragon Point and Speak is an introductory package meant for home users. It allows you to dictate in any Windows application at just a click of a mouse. You can change or correct by voice the last word you dictated and use the keyboard to make other changes. With the Standard version, you can format, edit and correct text completely by voice. The Preferred version is most popular as it allows you to play back your own dictation, use hands-free features and even dictate while you are away from the computer. For instance, a recorder compatible with the system can be used to take an interview, which gets transcribed when connected with the system later.

IBM has introduced a similar range called ViaVoice with three variants-Executive, Office and Home. The

Executive model provides the most comprehensive and personalized

vocabulary.

Going local





Most of the speech products are offered for all platforms including MS DOS, Windows, Windows NT, Macintosh, OS, DSP and telephony platforms. Dragon's products are available in seven languages: American English, British English, French, German, Italian, Spanish and Swedish. The main concern for India and other Asian countries is that the products are more applicable in the US and Europe. "We have realized that for total acceptance of the product it has to be made language specific. We are talking to HCL to develop a Hindi version. Japanese and Chinese are the other Asian languages being considered," says Vanderstoel.

As part of its localization efforts in India, IBM has commissioned a study that will enable the company to plan according to the needs of the domestic market. "We need to identify appropriate channels like speciality bookshops or peripheral manufacturers so that availability of the product is not a problem. An `extra-bit' of hand-holding is also required to make this technology popular," says Arora.

Dragon also works closely with OEM licensees and independent software vendors. Many companies like Articulate Systems, Digital, Genosoft, IBM, Microsoft and Seagate have integrated speech from Dragon Systems into their product offerings. The company aims at such integration in India as well. "If we want to replace the keyboard and mouse, it is important that we start supplying our software alongwith the PC. "This will increase the acceptability of our product," said Vanderstoel.

Although the technology is new, vendors are trying to provide products at an affordable price. While IBM's product is priced at Rs10,000, Dragon's cost ranges between Rs8,000 and Rs40,000. The company is targeting mainly the business and high-end home market, but IBM is also looking at journalists and publishers.

The downside





SR software has come a long way since its inception, but it has always been plagued by problems stemming from difficulties of understanding different types of voices (eg male v/s female), parsing sounds when people have different dialects (eg different accents), distinguishing between background noise and commands issued to the computer. Moreover, if SR has to work in real time, the software must have access to a large, fast database of known words and the ability to add more words.

Vanderstoel though admits that there are serious problems, he also hopes that they will be soon resolved. "We are trying to extend our product to various accents. At present, we are working toward a comprehensive English product." IBM is also working at an Indian English product. "We want the system to understand any user's command without even being trained or enrolled according to his voice. Earlier, enrollment was a rigorous procedure," explains Arora.

There are some areas were SR and NLP are not the best forms of human-computer interaction. For instance, mathematical equations or other logical languages are difficult to communicate via the spoken word. If a user wishes to convey some sensitive information to a computer, SR and NLP technologies dictate that the user speaks aloud, allowing anyone to overhear.

In spite of various drawbacks, Vanderstoel feels that SR systems are now mature enough for effective application. "It is only the attitude of the end-user which needs to change. Many people still view SR and NLP as some products from a science fiction," he said. "Before 1995-96, Dragon was mainly involved in research-specific solutions. But today, 80% of its revenue comes from product sales."

If the application difficulties can be minimized, SR and NLP could revolutionize the entire field of human-computer interaction. They will eliminate the need to understand the internal working of computers. Physical interaction will also be minimal. This means no more sitting before the computer and manually manipulating a keyboard or mouse.

Clearly, most people can speak much faster than they can type. If a user can convey an idea in four seconds that would otherwise take 20 or 30 seconds to type in, productivity can be improved. Perhaps, the greatest benefit that NLP will yield is the ability of a computer to understand a user's desire so profoundly that it is able to act autonomously on his behalf. The numbers are promising and vendors too appear optimistic. However, it will take some time before SR systems move beyond dictation and can be used for effective conversation with the computer.

Shweta Verma,





in New Delhi.

Back To Home size="-2" face="arial">| Back

To Content

Copyright Cyber Media India Ltd.



You must read the terms and conditions before
using this service