Hear and now

Build technology and "they will come" doesn't fly in the real world. Technology has to solve problems, be affordable and simplify our lives if it is to become widely accepted. It must be simple to use - and it's a real plus if it is invisible to the user.

Today, speech recognition lives up to these standards. Now is the time for service providers to make their customers' lives easier with the simple tool of speech.

Speech recognition has been around for more than a decade, but in the past year it has finally matured for three reasons.

First, computing power costs have reduced. Speech recognition is a very processor-intensive technology. The amount of power available for the money is doubling every year, making it much more cost-effective for developers and service providers alike. The result: faster and more powerful processors that are less expensive.

Second, the technology itself is more accurate. The industry has developed better algorithms, which leads to improved accuracy. In recent studies, in which consumers were asked to speak the names of companies into the phone, the voice recognition system recognized the name 99.4% of the time. That's 10 times more accurate than dialing. With 99.4% accuracy, users become very tolerant, which leads to the third reason: growing public acceptance.

Any technology is only as good as it is useful. With the improvements in today's speech systems, voice recognition technology is no longer reserved for technologically savvy road warriors. Today's systems go beyond a simple "Press or say 1" for this or "Press or say 2" for that. Now the latest algorithms allow for natural language. Figure 1 shows how the speech recognizer is programmed. The speaker merely has to talk normally and the system will pick up on the key words of the request.

For example, a talking Yellow Pages service could allow a caller to say, "Give me listings of Italian restaurants in Springfield." The system would home in on the words "Italian," "restaurants" and "Springfield" and provide the information so the caller doesn't have to navigate through several layers of menus. That kind of simplicity makes the service palatable to the public.

Speech recognition can be a tool for wireless and wireline service providers alike. In the wireless arena, this will be the year for speech because the service ante has been raised. Most PCS carriers have completed initial buildouts of their networks and are now sitting on a lot of coverage and capacity. Many are bundling what used to be add-on services, such as voice mail and call waiting, into basic service packages. Now carriers are looking for the next wave of services to differentiate themselves in this highly competitive market and to generate additional revenue. Speech recognition can bethat next wave.

Services that use the spoken word as an interface are especially attractive to wireless providers as a selling point to drivers. Most wireless users do other things while talking on the phone, but using a mobile phone while driving can be distracting. This is especially true when customers use advanced features, such as checking voice mail, initiating three-way calls or checking stock quotes. All these attractive features require a great deal of touch-tone manipulation. But voice-activated services help drivers keep their hands on the steering wheel and their eyes on the road.

Making it work in the network

A complete telephony-based speech solution consists of a speech processing server, a set of industry-standard PC boards and software to support advanced speech applications. For smooth operations, the speech server should integrate with intelligent network or call center platforms, as well as central office or customer premises equipment.

Service providers should look for solutions that fit into standard CO frames and can accommodate the most users with the fewest number of boxes, taking up the least amount of floor space to keep installation and maintenance costs down. Service providers should be careful of adjunct solutions, which may have trouble scaling up as the demand for service grows.

In addition, the adjunct-type systems do not integrate well into service providers' existing operations, administration, maintenance and provisioning systems. To position themselves well for the future - and thus improve both their competitiveness and their bottom line - service providers should look for highly reliable speech servers that scale to adapt to their growth needs.

The series of boards, which are usually integrated onto an intelligent network service node, include the speech processing board, which serves as the engine for recognizing natural language; the T-1 interface card, which links the application to the network; and the echo cancellation board, which improves the quality of voice connections and allows users to interrupt the system. To develop applications, service providers either work with systems integrators or develop specific offers in-house.

With the technology and the market ready, service providers can offer profitable new services that help make customers' lives easier. With the high level of accuracy that's now available, even services that have been around for a few years - such as voice dialing - are better than ever.

In the past, voice dialing never became widely accepted because it was cumbersome to operate. It took too long to create address lists and there were too many prompts, which didn't really save time over touch-tone menus. The user had to say the name being entered in the address book several times to train the system. Using systems based on the old technology, the user had to go through several layers of prompts, following a long scenario such as:

Prompt: "What do you want?"

User: "Call."

Prompt: "Call who?"

User: "John Smith."

Prompt: "At home or at his office?"

User: "At home."

Prompt: "Do you want to call John Smith at home?"

User: "Yes."

Prompt: "Calling."

That wasn't exactly easy to use - it was sort of like talking to a dog. Now with natural language interfaces, creating address books and making calls with voice dialing is as simple as speaking into the phone:

Prompt: "What can I do for you?"

User: "I want to call John Smith at home."

Prompt: "Calling John Smith at home. Say stop to cancel."

With the technology available today, voice dialing is realizing its potential for helping people save time. And the sky's the limit for developing all kinds of new services as well.

Blue sky services

As public acceptance for speech recognition grows, the number of creative services that a provider can offer flourishes.

A twist on voice dialing, for instance, is company dialing, which takes advantage of a shared list among a group of users. For example, a wireless service provider could examine the most frequently called numbers in a market, develop a shared list and offer a quick-call resource for its customers. The list could include the most popular pizza places, travel agents, daycare facilities, dry cleaners - any place that people often call from their wireless phones.

The service provider could generate even more revenue by charging the retailers for being included in their lists. Or it could create joint marketing programs with its top business customers. The promotional opportunities are vast and the users benefit from a tool that makes their lives easier.

There's a whole world of information access waiting to be tapped with voice recognition technology, from local weather and sports to roadside assistance and street direction services. One example is being planned by MovieFone, the nation's largest movie guide and ticketing service, known by such familiar local phone numbers as 777-FILM. Callers will be able to get movie listings and tickets faster and easier by simply speaking into the phone (Figure 2). Because today's speech recognition systems also have an excellent understanding of numbers, moviegoers also will be able to enter their credit card numbers verbally, rather than manually, to pay for tickets ahead of time.

Until quite recently, the idea of having an electronic personal assistant seemed either futuristic or merely for the wealthy or powerful. But today an automated secretary activated by speech is a feasible offer for service providers because it's realistic for all kinds of customers. A system that screens calls, takes messages and finds the subscriber when an important call comes in becomes marketable when it's easy to operate with simple intuitive language instead of awkward commands.

Using a personal secretary to handle messages is attractive for both wireline and wireless customers because it helps users become more organized and productive. With a personal assistant, users can get critical messages immediately rather than having to navigate through all the others first. A personal assistant also is helpful when a driver receives another call while talking on the phone. The personal assistant announces who's calling, so if the user doesn't want to take the call - or pay for it - the call can be sent to voice mail.

When a service becomes so routine that users don't have to think about it, they will use it more often. And of course, the more a service is used, the more profitable it is for service providers.

It's important for service providers to remember that speech is a means to an end: helping make their customers' lives easier. There's a tremendous potential for this market. Speech recognition is at a point where it's ready to be taken seriously by service providers that want to generate additional revenue and serve their customers with interesting new applications. The arrival of speech services is no longer a question of if, but a question of when. And the when is now.