Marchex, a mobile advertising analytics company, today announced the results of a product test, independently verified, of its recently launched Marchex Speech Analytics solution. Enterprise Technology Analyst Frank Ohlhorst put Marchex Speech Analytics head-to-head with the recognized leader in automatic speech recognition (ASR) systems, IBM’s Watson, and concluded that Marchex is setting the standard in speech analytics.
Marchex Speech Analytics enables enterprise and mid-sized companies to derive actionable insights from inbound phone calls from consumers to their business, to improve media spend and sales operations, and help convert more callers into customers. The recently launched technology is gaining insights daily and continuing to evolve and expand its capabilities. To validate its ascendency in the customer to business call analytics space, Marchex had an independent third party directly compare its capabilities against IBM’s Watson.
“We know this is only a sub-set of Watson’s full capabilities, but for companies looking to evaluate call data and analyze consumer to business phone calls, Marchex Speech Analytics proved superior,” said Ohlhorst. “That said, there are still elements that can be compared between Marchex and Watson, which include the accuracy of how voice is processed and ultimately transcribed, and Marchex outperformed Watson from that perspective.”
Marchex Speech Analytics and Watson were evaluated through a series of tests to directly compare their capabilities, identifying two key metrics to measure speech technology performance: RAW word error rate and perceived word error rate, ensuring fair testing for both systems. RAW word error rate measures the number of words that are inserted, deleted and substituted in order to discern overall accuracy of a transcription, while perceived word error rate normalizes some words and spelling that are really the same, but in the RAW case would appear as an error. For example, if a human transcription recorded a word as “alright” but the machine transcription showed “all right,” the RAW system would consider that to be an error while the perceived system would not.
Marchex Speech Analytics outperformed and achieved a lower rate of error than Watson in both categories:
Watson achieved a RAW word error rate of 21.1%, compared to Marchex’s RAW word error rate of 15.7%
Watson demonstrated a perceived word error rate of 19.4%, while Marchex achieved a lower rate of 14.6%
“Accuracy is critical for those seeking actionable insights from within a phone call,” said Jason Flaks, Senior Director of Product and Engineering, Speech Analytics at Marchex. “These test results show that our technology can deliver the accuracy to not only understand what happens on phone calls at scale, but also make informed decisions on how brands should be interacting and engaging with their customers.”