Results from Word Error Rate (WER) testing reveal newly launched proprietary speech recognition service outperforms tech giants

Rev.com launched Rev Automated Transcription, a service that creates affordable, easy-to-follow transcripts that power productivity. The new service complements the company’s established leadership in human-powered transcription and adds functionality to Rev’s convenient platform. The automatic speech recognition (ASR) engine powering the service transcribes speech-to-text in minutes and exceeds 90 percent accuracy on files with clear audio, beating Google, Amazon and Microsoft in tests.

Read More: Ericsson Extends Global Supply Chain With Company’s First Smart Factory in the U.S.

“We realize there are times when businesses and individuals may prefer an automated transcript over a human transcript. We’re thrilled to better serve customers by offering both options in one convenient platform,” said Jason Chicola, Co-founder and CEO of Rev.com. “Our scientists have built one of the best speech recognition engines on the market, empowering users to store, search and capture insights from the spoken word. We hope to become an integral part of people’s lives while liberating users to focus on their skills and let Rev drive efficiencies for them.”

Established in 2010, Rev has become a trusted provider for fast, accurate, and affordable audio and video-to-text needs. Unlike others in the voice market, Rev built its ASR engine from the ground up, training it on more than fifty thousand hours of accurately transcribed audio samples. Results from a recent Word Error Rates (WER) test on a collection of public podcasts illustrate the accuracy of its ASR engine. Rev’s service yielded the lowest WER compared to major providers in this space. More information on the test’s methodology can be found here.

Read Also: Trax Acquires Shopkick to Bridge the Gap Between Mobile and Physical Worlds of Retail

In addition to excelling on word accuracy tests, Rev Automated Transcription captures highly accurate punctuation and can distinguish between speakers, creating easy-to-follow transcripts. Rev achieved industry-leading results by developing software that leverages large quantities of accurately transcribed audio data and applying advanced deep learning methods. Rev also uses Natural Language Processing (NLP) techniques to derive meaning from the content of colloquialisms, abbreviations and acronyms, producing more accurate results without a human.

Rev Automated Transcription is an ideal tool for capturing meeting notes, phone calls, interviews, film production and script editing, lectures and more. Key benefits include:

  • Play and edit transcripts alongside audio or video playback as well as adjust playback speed to review at one’s own pace.
  • Identify exactly where the speech recognition engine may not have accurately transcribed something with color confidence scores so users know where to focus on editing.
  • Customize time stamps at no extra cost.
  • Identify different speakers and easily edit names.
  • Search, comment on, organize and share transcripts from Rev’s desktop application.
  • Export final drafts into Word, TXT, PDF, srt or.vtt file formats.

Read More: Afilias To Acquire .GLOBAL Top Level Domain

“I was skeptical about trying an automated service. In my experience with voice technology, it’s not consistent or very reliable,” said Jon Reed, Co-founder of Diginomica. “However, I’ve definitely found a useful role for Rev’s Automated Transcription service in my transcription mix. If the sound quality of the recording is good, there are times when I am blown away by the accuracy they have achieved — even with highly technical audio content. I will always have the need for a word-for-word account from Rev’s human service, but I can see myself making more use out of automated transcripts because this service is fast and more affordable.”

As audio and video become the language of the internet, automated transcription is the fastest and easiest way to pull insights from these formats. Rev Automated Transcription saves users a great deal of time and resources by turning around files 99 percent faster than its human service at a fraction of the price.

Read Also: The Linux Foundation Will Host the Federated AI Enabler to Responsibly Advance Data Modeling