Phonetic recognition systems now break down spoken words into small fundamental sound units called phonemes. When compared to word-based recognition systems, phonetic recognizers enable increased accuracy in understanding larger vocabularies. Language modeling technology is also used to heighten accuracy by comparing recognized sounds to a list of usage rules or constraints to determine the probability of one sound following another (see Figure 2).

Figure 2. Phonetic Speech Recognition

Speech solutions are now enabling the development of IVR applications that go beyond rigid touch-tone interface models to exploit the navigational flexibility offered by natural language processing. To aid recognition and reduce complexity, an "n-best list" of the most likely word or phrase matches to the spoken utterance and associated confidence scores are then created and interpreted by the system. Natural language recognition and advanced user interfaces that conduct interactive dialogues with users in order to complete transactions are driving the creation of the most versatile and robust applications ever developed for the IVR industry.

Why Are Many Customer Service Organizations Incorporating Speech Recognition into Their IVR Systems?

Many factors are driving the emergence of speech as the IVR user interface of choice for today and tomorrow. The first is spiraling labor costs. The cost of employing live customer service agents is rising at the same time that organizations are facing increased pressure to reduce the cost of serving customers. When an automated call-processing solution must be employed, a speech-enabled IVR application increases caller acceptance because it provides the friendliest and fastest self-service alternative to speaking with a customer service agent. Speech solutions also create new opportunities to automate transactions that are too cumbersome to complete using a DTMF interface, such as bill payment or stock trading.

Higher call volumes make the addition of speech recognition more cost-effective. In this era of mergers and consolidations, customer service organizations are often large enough to realize significant cost savings and a quick return on their investment in creating and deploying speech-enabled IVR applications.

Recognition of larger vocabularies is faster and more accurate with today's increased processing power, and hardware and software costs are now lower. Today's IVR systems are more robust and have the processing power required to support the latest speech recognition technologies. In addition, speech solutions for IVRs can be less costly and faster to implement than Web-based solutions.

The Gartner Group reports that an added advantage of voice recognition is the ease of adding verbal security steps to reduce risk with minimal customer impact. Speech recognition eliminates the need for touch-tone keypad entry and makes it easier to ask and answer traditional security questions, such as social security number or personal identification number (PIN), so that customers are not as bothered by them as they are on an IVR.

Speaker verification technologies also promote enhanced security by testing vocal characteristics of spoken utterances to verify the identity of the speaker. Voice print technology is now available that seamlessly verifies the identity of the speaker while it concurrently recognizes conversational language. Applications for this technology are especially attractive for providers of financial services and telecommunications because the information exchanged to complete their customer transactions is often valuable and therefore subject to fraud.