Speech Technology Magazine

 

Marchex Updates Speech Analytics Interface

Marchex has updated its Call DNA interface to accept more inputs and feedback.
By Phillip Britt - Posted Oct 26, 2017
Page1 of 1
Bookmark and Share

Marchex has updated the Call DNA user interface for its speech analytics solution. The interface now accepts user input on artificial intelligence-predicted outcomes, enabling Marchex to continually improve the overall accuracy of its speech recognition technology.

"We work with a large number companies that are trying to make sure they optimize the number of customers they can convert," says Jason Flaks, Marchex's senior director of products and engineering. "If you want to optimize your sales pipeline, the best way is to really understand the call. We do that by transcribing them and providing them to the customers."

Call DNA classifies the outcome of the calls, identifying which ones are sales-related and which are service calls, as well as how successful the sales calls are. So the technology helps companies determine ways to improve the calls, Flaks explains.

U.S. conversational speech is a very large domain, Flaks added. "By being super-accurate, we catch the tail of speech. There are words that some might think are irrelevant but are critical to understanding the conversation."

For industries that rely heavily on phone calls from consumers to drive appointments and sales, AI-powered conversation technology offers a better understanding of the customer experience and the most accurate view into exactly what happens during every call at scale, according to Flaks.

"When General Motors rolled out Marchex Speech Analytics, we were amazed at the insights it provided," said Jim Bechtell, vice president and general manager at Somerset Buick/GMC, in a statement. "We are using this product every day to better understand our sales agent behaviors and to improve our customers buying experience."

The Call DNA update follows a significant enhancement to the Marchex Speech Analytics dataset, which promises greater transcription accuracy by reducing the word error rate by 15 percent. The increased accuracy of conversational speech recognition enables marketers and sales teams to gain deeper insights from inbound phone calls, which can then be used to optimize media spend and sales operations.

"We are consistently training our system with enormous sets of data," Flaks said.

According to a recent comparison in Gigaom, Marchex's speech analytics technology was found to be more reliable than IBM's Watson. While traditional call analytics solutions typically rely on training data in isolated data sets, Marchex's conversation analytics platform leverages its speech recognition and Call DNA technology to aggregate and analyze tens of millions of utterances from millions of unique phone calls for language model training.

"The English vocabulary consists of more than 125,000 words, which can be put together in trillions of ways," Flaks said. "Accounting for all of those possibilities is difficult but critical for brands looking to understand what happens on their calls with consumers."

Marchex released its purpose-built speech analytics technology and tools earlier this year. The company purports to have the highest accuracy rate when identifying voice-patterns during customer interactions with companies.

As a recent industry study indicates, customers who call businesses are more valuable prospects than those who don't. The report found that phone customers convert faster, spend more, and have higher retention rates than customers who contact companies via other channels, such as in-store or online. This further highlights the need for businesses to leverage conversation analytics technology to better understand what is happening during customer journeys so they can convert more callers into customers.


Page1 of 1