Marchex’s AI-Powered Speech Analytics Sets New Standard, Approaches Human Levels of Accuracy in Consumer to Business Conversations

By STS News Desk On Feb 14, 2019

Marchex’s AI-Powered Speech Analytics Sets New Standard, Approaches Human Levels of Accuracy in Consumer to Business Conversations

Driven by investments in data science and deep learning technology, Marchex delivers best-in-category insights on customer conversations

Marchex, a leading provider of call analytics that drive, measure, and convert callers into customers, announced that is has reached a new industry standard in automatic transcription accuracy for consumer-to-business phone conversations, driven by enhancements to and investments in its proprietary speech analytics solution.

Marchex’s speech recognition functionality was put to an independently verified test against several other commercially-available solutions, based on a series of factors to directly compare capabilities. Using the same sample audio data from the business conversation domain, the test measured the word error rate (WER) which tracks the number of words that are inserted, deleted or substituted in order to discern overall accuracy of a transcription. The internal evaluation, which was verified by enterprise technology analyst Frank Ohlhorst, found that:

Marchex achieved an overall WER of 8.4 percent, which is a 20 percent improvement since 2017, and considerably better compared to other voice recognition systems.
Marchex’s speech analytics solution performed 38 percent more accurately than IBM Watson, 37 percent more accurately than Azure Cognitive Service API, 13 percent more accurately than Google Cloud Speech-to-Text API and more than 11 percent more accurately than Amazon Transcribe API.
Marchex speech analytics technology is nearing parity to human accuracy, which is roughly a 5 percent WER¹.

For industries with a significant reliance on phone calls to drive sales, such as the automotive, travel and home services industries, AI-powered conversation analytics technology is critical to delivering sales enhancing personalized solutions. Accuracy plays a significant role in being able to automatically pinpoint the reason why a customer is calling, the outcome of the conversation, and the ability to classify calls in real-time.

Read More: Twilio Completes Acquisition Of SendGrid

“Marchex takes a multi-pronged approach to resolving the issues associated with human to human interactions in the enterprise business space,” said Frank Ohlhorst, technology analyst. “Processing accuracy is a viable benchmark and Marchex was able to outperform other solutions, especially where accuracy was concerned.”

Marchex invested in several key enhancements to its speech analytics technology, including additional deep learning capabilities, which have expanded both the accuracy and capacity of consumer-to-business conversation analysis. In addition, Marchex has invested heavily in internal resources, including the launch of the Marchex Innovation Development (MIND) Lab and the addition of several new data scientists. These investments represent Marchex’s ongoing dedication to lead and innovate within the speech analytics sector.

“In recent months we’ve prioritized resources and targeted investments to further advance our speech technology,” said William Li, vice president of engineering at Marchex. “With our data science team leading our MIND Lab, Marchex is delivering meaningful insights that help businesses accelerate sales and deliver a better conversational experience for their business.”