customer service: utility stylea primer on speech recognition

Penni McLean-Conner, P.E. vice president customer care, NSTAR

As we were finalizing our purchase agreement for a new speech-enabled, hosted interactive voice response (IVR) solution, I reflected on the number of new terms, concepts and acronyms associated with this technology. If your utility is considering a speech-enabled IVR system, understanding terms and acronyms associated with the technology will help you evaluate speech recognition options and systems more effectively.

utilities widely adopting IVR

Speech recognition has been around since the early 1990s, but in recent years it has matured significantly. Today’s speech recognition systems respond to callers in a conversational manner and can decipher a variety of languages, accents and dialect. And customers respond to speech recognition systems. Research indicates that 47 percent are much more satisfied with a voice recognition application than touchtone. Utilities that implement automated speech recognition do so not only to increase the percentage of calls handled by self service, but also as a strategy to increase customer satisfaction.

Utilities have widely adopted IVRs. According to a 2005 Chartwell report on IVR applications and speech technologies, more than 63 percent of utilities use this technology to handle some types of calls. Upgrading the utility IVR is a trend in the industry and many utilities choose to incorporate speech recognition in their IVR during re-engineering efforts.

One of the key components to speech-enabled IVRs is the design of the voice user interface (VUI). The VUI includes the call flow of the speech recognition system and the voice and persona of the system itself.

The objective of VUI design is “not about fooling callers into believing they are talking to a live call center agent, but rather making callers forget that they are talking to a machine,” wrote Tom Houwing and Paul Greiner in “Design Issues in Multilingual Applications.” VUI designs today offer the options of natural language and directed dialog.

In their book “Call Center Technology Demystified,” Lori Bocklund and Dave Bengtson define natural language as technology “used in speech or text recognition that recognizes what is being said or requested through free-form communication.”

The benefit of natural language is eliminating a series of options a caller must listen to before selecting his or her choice. A natural language interface could begin with, “Thank you for calling (your utility), how may I help you?” Well-designed natural language interface can reduce the call handle time within the IVR by more quickly routing a customer to the desired service.

Bocklund and Bengtson define directed dialog as a “speech recognition approach that recognizes what is being said based on guided or structured interactions.” In this interface, a caller will be offered a series of options such as, “Would you like to pay by check or credit card?” The caller will respond by voicing one of those two options.

other terms to know

There are several other terms used repeatedly in speech-enabled system discussions.

Utterances: The actual words and phrases customers say when they first speak to a call center representative.

Utterance collection: The process of recording thousands of customer responses to an open-ended prompt. The utterances are used to train the speech-recognizing systems and to map call flows within the business process.

Intent determination: The process of understanding a customer response to an open-ended prompt and mapping it to a specific business process; also referred to as topic spotting.

Disambiguation prompts: Follow-up prompts to open-ended prompts, used if the caller’s first response does not contain enough information to map it to a specific business process. For example, if a customer’s first response is, “I have a billing question,” the disambiguating prompt could be, “Are you calling for your account balance, a high bill or where to send your check?”

Call containment rate: Calls completed in the application divided by the total amount of calls offered to the application. Call containment rates are a measure of the effectiveness of the speech enabled self-service applications.

Speech-enabled IVRs are a powerful tool utilities can use to increase customer adoption of, and satisfaction with, self-service applications on the IVR. At the kickoff of our speech enabled IVR re-engineering project, I reminded our project team of this, and that one of our top customer complaints is the complexity of our IVR. The great news is that once completed, this project and others like it will dramatically change and improve the customer experience.

Penni McLean-Conner, a registered professiaonl engeneer, is the vice president of customer care at NSTAR. Her first book, “Customer Service: Utility Style,” has just been published by PennWell Books.

Previous articlePOWERGRID_INTERNATIONAL Volume 11 Issue 4
Next articleSuntech Power receives UL cert for photovoltaic modules

No posts to display