Visa International is betting that “v-commerce,” the heinously-named catch-all term for telephone transactions enhanced by speech-recognition, is a big part of their future.
Since 1995, Visa International has been an investor in Nuance, and has participated in pilot programs to add automated speech to cardholder transactions — basic things like card activation, card replacement, and travel planning.
These are things that, like all good IVR apps, don’t need an agent for the basic, introductory information gathering stage of the transaction. Only when things get more complicated, or the consumer gets confused and tries to bail on the automated system, does an agent really become necessary.
Now Visa and Nuance are working on developing apps for use by the member banks for customers to self-serve over the phone. It’s got some important names from the call center field, but it doesn’t have any speech companies other than Nuance. Nuance’s speech rec is as good as anyone’s, but it’s not the only one, so the impact of an association like this is in the power of the end-users, like Visa, to push a particular technology.
Several v-commerce applications are currently available for Visa member banks including speech banking and automated bill payment.
Another really interesting real-world application of speech rec is the installation that went into Ameritrade (the brokerage firm) in early 2000. Ameritrade decided to take their existing (and recently installed) speech rec front-end and expand its capacity.
Ameritrade’s system lets their brokerage customers check their accounts and act on their investment decisions via telephone using natural speech recognition. The speech-enabled system was introduced to Ameritrade customers on March 10, 2000, and handled more than 650,000 calls in its first nine trading days. That huge response was a major factor in Ameritrade’s recent decision to increase the port capacity of its InterVoice-Brite call automation system. The expansion is planned for this month.
Since the speech-enabled system was implemented, Ameritrade’s call completion volume has significantly increased. The system currently handles an average of more than 85,000 calls on trading days. More than 40% of callers are already opting to use the speech recognition capabilities. This self-service transaction option gives Ameritrade customers a faster, more convenient method for rapid stock transactions while enabling the company to increase the efficiency of its call center by reducing call wait times and freeing agents to process more complex customer service requests.
The system isn’t pure-speech only; rather, their call flow involves an interesting hybrid of traditional user-entered touch tone digits and speech input. To use the self-service stock trading system, callers enter their account code and personal identification number, using traditional keypad entry. Then, the system asks for which stock the caller would like a transaction. Rather than using a tedious touch-tone entry method, callers speak the company name or stock symbol in a natural voice.
It runs on InterVoice-Brite’s OneVoice platform and uses technology from SpeechWorks. The system’s vocabulary exceeds 60,000 words and even recognizes popular stock nicknames, such as “Big Blue” for IBM. (Though I believe you’d have to be really wanting in common sense to actually try to trade stock using that kind of nickname.)
InterVoice-Brite has also put in speech-enabled stock systems for DMG & Partners Securities, Lim and Tan Securities, Keppel Securities in Singapore, and Hyundai Securities in Korea.
The system will be able to interpret more than 80% of first and last names in the United States. That’s going to make the system more viable for applications like health insurance benefits verification, travel reservation verification and cancellation, and inquiry applications where callers are asked to leave their name and number for an informational callback.
New multilingual capabilities simultaneously support two or more languages on a single system. These multilingual capabilities will enable application developers to create self-service applications in 10 different languages. Callers can also respond to prompts in their own dialect depending on geographic location and regional demographics.
Languages supported now include English (US, UK, Australian and Singaporean), Spanish (Latin and US), French (European and Canadian), Chinese (Mandarin) and German. The system features a vocabulary with more than 70,000 words and supports an increased number of ports to support higher call volumes. Additional tools include custom vocabulary development, industry-specific grammar libraries and a self-tuning feature.
No comments:
Post a Comment