Startup Uses AI and Human Augmentation for Video/Audio Transcription
- By Dian Schaffhauser
- 06/04/18
A startup based in Israel has raised $11 million to expand the growth of its solution for doing artificial-intelligence-powered transcription. Verbit technology, according to the company, will be helpful to schools in addressing the Americans with Disabilities Act (ADA) regulation.
The transcription process for videos and audio recordings typically relies on a combination of approaches: fully automated transcription, which tends to produce partially accurate results, and/or manual transcriptions, which require a much longer turnaround. Verbit has developed Verbatizer, a transcription system that uses a combination of AI technologies for automatic speech recognition algorithms and human-augmented refinement. The corrections made by human transcribers are fed into the Verbit algorithms through machine learning technologies to continuously improve the formulas.
Verbit said it has already been tested by higher education and major e-learning customers, including London Business School, Florida International University, Stanford University, Coursera and Udacity.
In a description of the service, Verbit said an educator can submit video content just moments before the class begins to obtain a transcription with an average accuracy rate of about 90 percent. Within a couple of hours of the end of class, the transcriptions will have a 99 percent accuracy rate. This kind of turnaround can, the company suggested, be a great aid to students who have submitted captioning, interpreting and transcription requests.
"We already generated millions of dollars in revenue, but we've only just begun. With this funding the company set aggressive growth targets and significantly expanded product capabilities," said CEO Tom Livne in a press release. The company plans to double the number of employees in the coming year on its way to realizing its vision: "To make the world's verbal content accessible and searchable."
A comparison of Verbit's transcription capabilities against the services delivered by Google Docs and Watson are available online here.
About the Author
Dian Schaffhauser is a former senior contributing editor for 1105 Media's education publications THE Journal, Campus Technology and Spaces4Learning.