The accuracy and speed of speech recognition is greatly improved thanks to faster computers and improved software. Speech recognition uses specialized computers (digital signal processors) to capture utterances, compare them to particular databases and translate them into digital representations of for example, a name in a directory. The software is sophisticated enough to "learn" regional accents and pronunciations and remove background noise.

How do organizations use speech recognition?
Organizations are using it for automated attendant and internal dialing. For example, if a remote employee or caller knows a name but not the spelling they can easily speak the name to be connected to the called party. Thus, callers only need to learn the main number. These systems are also used for extension to extension calls to decrease information calls to company operators.

Additionally, call centers use it for routine queries such as train and airplane schedules, prescription refills and order tracking to decrease the number of "menus" callers must navigate.

Speech recognition in voice mail enables speech commands instead of touch-tone to for example, skip ahead or backup while listening to messages.

Why do call centers use speech recognition?
A major challenge in call centers is staffing for uneven call volumes. For example, an ad in a magazine can trigger exceptionally high call volumes. Per Bob Brennan, IT Director for Specialty Catalogs, one of their ads prompted 30,000 calls for catalogs. Based on past history and marketing predictions, the organization had expected 8,000 calls. Specialty Catalog is interested in a system to handle these spikes.

What equipment is required?
In call centers, speech recognition software is installed on integrated voice response systems (IVRs). IVR equipment connects to databases of for example, airline schedules.

Automated attendant and voice mail speech recognition software is installed on separate computers or integrated into voice mail systems.

What about pricing and offerings?
Prices vary starting at about $20,000 to $35,000 for a four port system supporting automated attendant functionality for about 400 users. A Locus Dialogue system for 100 directory entries costs $8,000. Avaya, Parlance Phonetics and SpeechWorks International sell systems directly. Nextira offers LocusDialog systems.

Providers such as NetByTel in Boca Raton and Shoptalk in Chicago host speech recognition services so call centers can try speech recognition without a capital investment

Enhancing user acceptance
The key to any successful implementation is the user interface. For example, how long do callers have to wait until they can speak a request? Can repeat callers interrupt the prompts to speak their request? Finally, directories need to be kept up-to-date. New names and nicknames need to be added so callers aren't frustrated by unrecognized entries. Some systems are updated automatically using links to existing PBX, Human Resource or standards based directories. Finally, call centers need to design menus suited to their callers. Testing the interface before it goes live is critical.