Nuance Installs NLP System for Nordic Telco

In recent years, there’s been a great deal of interest in natural language IVRs, and call centers have increasingly attempted to incorporate this technology, despite its current limitations. It was in this spirit that Nordic telecommunications company Telenor replaced its directed dialogue IVR with the Call Steering application from Nuance Communications. The application runs on the Dolphin HotVoice platform, integrated with the Nuance speech engine.

"The Call Steering system from Nuance efficiently provides the required service in minimal time, thus increasing customer satisfaction, which was the main driver behind the project," Lynda Kate Smith, vice president and general manager of the Nuance Care business unit, said in a press release. She went on to cite the solution’s 80 percent routing success. "Nuance natural language speech solutions handle more than half a billion customer care calls annually around the world," she continued.

The system, like most solutions of its ilk, uses natural language only for the initial How may I help you? prompt, then routes callers to the proper department. It’s a combination of natural language and directed dialogue that’s become popular for call center solutions. "Without it, users have to go through the 20 questions and that takes a lot of time and users don’t like that," says Larson Tech consultant Jim Larson. "It’s serving a real need and it’s been quite successful."  

The interface was initially developed by AT&T and now other companies, such as Nuance, have adapted systems of their own. The development of such an application is remarkably complex, which limits its scalability.  To implement a natural language approach for a single question, a developer must capture thousands of answers to the initial prompt, then determine what the user wants with each answer.  

"There’s a fair amount of manpower trying to get these things started," Larson adds.  Following collection of the data, developers have to create what’s called a statistical grammar to recognize key words in a phrase and route them to the appropriate department.  

The use of this How may I help you? interface is already increasing, though only a fraction of IVRs currently use it. Yet user demands for higher customer care standards might spur a proliferation.

"There’s lots of room for growth in the call routing area and I think it will continue to grow," Larson says. However, he does not anticipate a natural language interface will expand beyond the initial prompt. "It’s just too much work for the developers to create these large databases of utterances and then go through and mark them."  Asking a question such as How may I help you? already constrains what a user can say, crucial for the IVR’s success. While such a system works well, it only works well for very specific questions; in this sense, the natural language system isn’t a true natural language system.  

"It’s actually a fairly niche-y kind of usage, but it’s been very successful there," Larson says.  "Can it be used in other areas?  This is the only area I’ve seen it used successfully."