Verbatim-VR has developed technology that enables users of voice recognition software to correct recognition and pronunciation errors and have those corrections made available to all users within their specific market segment.
Users of the solution can perform audio editing, text correction, post editing, and more. Once the editing is completed, the individual user's voice recognition error corrections are immediately followed by updating the business-sector knowledgebase. This updated Shared Knowledge-Base contains each user's cumulative post-error-corrections for different pronunciations of words, thereby improving voice recognition accuracy. Each user receives, within a matter of hours, a totally error-free verbatim text copy of what he/she actually said during a specific voice recognition session.
"We extract information about each session, sentence, and word," explains Bill Drewes, Verbatim's CEO. "We have a relational database that updates the vocabulary and language dictionaries for everyone at that specific job site. Every correction updates across every dictionary entrance for it."
"No one wants to do error correction," Drewes says. "So we make the error correction available for everyone."
According to Drewes, business-specific voice recognition products covering fields like law, healthcare, and financial services, typically have hundred of thousands of individual entries in their language and vocabulary dictionaries. Most products typically encounter voice recognition error rates of 7 percent to 10 percent, "and this problem just doesn't go away."
Verbatim's solution, though, "will naturally lower this error rate significantly," Drewes says.
Drewes calls Verbatim-VR "an important milestone in voice recognition" that will transform it into the productivity tool it was meant to be.