Voci Releases Speedy Speech Recognition

Voci Technologies has launched V-Blaze, a commercial speech recognition appliance that converts audio files to text at a rate of 50 terabytes of data a year on a single appliance.

The speaker-independent appliance transcribes all the words of conversations into text, and also offers keyword spotting or scanning for instances of user-specified words of interest.

Voci’s proprietary application-specific architecture is implemented with computers from Convey Computer that run a standard enterprise edition Linux OS on an Intel Xeon processor. State-of-the-art reconfigurable Xilinx Field Programmable Gate Arrays (FPGAs) power the Voci speech accelerator. This approach liberates speech recognition from the constraints of software-only solutions.

Input to V-Blaze is standard audio file formats (WAV) or live audio streams, and the system outputs a full text transcript. Standard interfaces combine with a Web monitor interface and output can be incorporated into existing text analytics environments. An API is included for configurations requiring tighter integration.

V-Blaze is available now, at a starting price of $50,000 per year for a 3-year subscription license, which includes maintenance, training and a North American English language model and vocabulary. Spanish and other languages are being added. Perpetual licenses and hosted on-demand use models are also supported. 

One of the first clients of this new appliance is Mindshare Technologies, a provider of enterprise feedback management solutions. Mindshare will rely on the Voci appliance to convert customers' voice comments from surveys into text, and then analyze and monitor the text in the Mindshare Text Analytics Suite. Mindshare performs more than 30,000,000 surveys a year with millions containing voice comments. Before Voci, Mindshare relied on human transcription to convert a fraction of spoken input into text so it could be analyzed.

"The most convenient way for a consumer to give feedback is through a short, open-ended answer about their overall experience," said Kurt Williams, chief technology officer at Mindshare. "A single comment can have 70-90 insights about their experience, which is significantly more than we could ever get in a short survey."

Williams continued, "Once we saw the speed and accuracy of Voci, there was no question that Voci offers the best way for Mindshare to integrate speech recognition into our text analytics solution."

SpeechTek Covers
for qualified subscribers
Subscribe Now Current Issue Past Issues