• Investment World
• eWorld
• Catalyst
• Mentor
• Life
• Canvas
• Praxis
• Urban Pulse
• Brand Quest

Pankaj Vaish

Here's more on speech recognition technology in the use of computers, which holds out much promise.

IN the pursuit of exploring more applications of the microchip, research has focused on how devices that use these chips can respond to voice commands of the user. The idea of the use of `voice commands' in a man-machine-interface holds tremendous promise for civilisation. This entails enabling devices to respond to spoken words i.e. recognising speech.

Speech Recognition is the use of computers to hear and understand spoken words. This technology enables users to communicate with computers and other speech-enabled devices by voice.

The most promising use of speech recognition technology in business is to allow users (customers) to interact with a computer over the telephone. A typical speech recognition application allows the user to answer questions and provide information using a normal speaking voice. For instance, a customer may use the telephone to get the latest traffic reports, make travel reservations, check on the status of an order, or hear a list of movies at a local cinema — all by telephone and all without the need of a live operator. Nowadays, one of the most elementary uses of speech recognition technology is the use of `voice dialling' in cellular phones. We have reached the point where any product with a user interface will soon incorporate a voice user interface.

The speech recognition market has always held a widespread and intuitive appeal. Users want to communicate with their products in the same way they communicate with each other. They want full-featured products that are easy to use. Though these concepts are eye-catchy and attracting, the markets have not yet attained their potential. Few success stories exist in the speech recognition industry today. The story of the embedded markets is also no exception. Many of the promising markets within the embedded speech recognition industry pose huge challenges. However, interestingly, the market continued to comprehend speech recognition products in 2001 as a means of lowering costs, providing greater customer service and launching new products. The use of standards such as Voice eXtensible Markup Language (VXML) and Speech Application Language Tags (SALT) has opened up vast new markets and opportunities. Given these factors, speech recognition will very soon find its way to every office and home.

For instance, companies have invested heavily in human-powered call centres or DTMF (touch-tone) IVR (Interactive Voice Response) systems. But why should they adopt speech recognition? Well, it's due to the fact that it can save money. In most cases, the operating costs of a speech recognition system are about 10 per cent of a comparable human-powered call centre and a half to one-third the cost of touch-tone or Web systems.

No operator training is required, and turnover is never an issue. Speech recognition-powered call centres are available 24/7 and automatically expand to handle peak loads and off hours. Shorter call times mean lower long distance charges. It also improves the `customer experience'.

In some parts of the world, 80 per cent of callers encountering a touch-tone menu system will quit or elect to transfer to the operator. With speech recognition, this can drop to 20 per cent or below.

Easier site navigation as there is `No' numerical menus.

Shorter call times: The easy navigation and elimination of hold time makes the average call shorter than with either a human operator or a DTMF system.

Very fast service for experienced users. Well-designed `natural language' systems allow experienced users to move very quickly through to their desired service or product.

Environmental constraints

However, the implementation and use of speech recognition throws up several challenges, such as environmental constraints.

Speech recognition applications require end-users to speak audibly, possibly for extended durations. The main issue quite unique to speech recognition is that poor acoustic conditions, such as high noise levels or high echo levels, can compromise recognition accuracy.

Variability of speech patterns

Different people speak the same language differently and even speak the same word in many different ways. Interpreting speech variability has led to the development of complex pattern analysis. Understanding various accents, natural pauses, speaking rates, and changes in volume is a complex and difficult task.

Using it for the better

End-users must be familiar with the correct use of the particular software and headset, microphone, or recorder. They must also be able to speak thoughts clearly and fluently in a language supported by the speech recognition application. Also, designing systems that are powerful enough to understand and respond to continuous speech requires a large amount of processing power. When a person speaks at a natural rate, it is difficult to distinguish which sounds are associated with specific words. Very few speech-recognition applications are able to accurately determine the meaning of words. For example, the phrase `to recognise speech' could be misunderstood to be `to wreck a nice beach.' Users do not naturally speak - with - pauses - between - their - words. As a result, processing phrases in real time as they are naturally spoken continues to be a major challenge.

The industry now has to tackle these challenges amongst several others before speech recognition becomes accessible to the common user.

(The author is GM - Telecom Practice, HPS.)

Graphics by K.B. Jawaharr

Article E-Mail :: Comment :: Syndication

Stories in this Section
More value, more dough

Copyright © 2003, The Hindu Business Line. Republication or redissemination of the contents of this screen are expressly prohibited without the written consent of The Hindu Business Line