Intelligent Voice Releases JumpTo Speech-to-Text

Intelligent Voice, providers of compliance and speech technology solutions, has released JumpTo, a speech-to-text appliance based on the new NVIDIA Tesla K80 GPU accelerators. Working with Boston Limited's High Performance Computing, Intelligent Voice, has created a series of 1, 2 and 4u servers to meet the most demanding speech-to-text applications.

The JumpTo system takes audio files, particularly telephone calls, and turns them into text. It then automatically indexes the resulting text using its Complexus technology to create bookmarks—key phrases that exist in the audio file. These not only serve as a visual snapshot of what is in the audio file but act as navigation to allow the user to jump to the point in a file where the phrase was said.

"Traditionally, speech-to-text has been seen as slow, and therefore difficult to use for compliance and monitoring purposes. With efficient capture and processing, we can now process the daily telephone calls of thousands of traders on a single appliance," said Nigel Cannings, chief technology officer at Intelligent Voice, in a statement. "We can also now offer ultra-high speed conversion of archival video and audio material with the added benefit of the JumpTo bookmarks, allowing instant access to the key parts of the material. This is, we believe, the fastest speech-to-text box commercially available. We have invested a lot of time and research in making speech technology practical, and are constantly striving to improve the speed and efficiency of our systems."

"By utilizing Supermicro-optimised SuperServers supporting the new NVIDIA Tesla K80 GPUs, we have seen a huge gain in the areas of computational chemistry and molecular dynamics, energy exploration, and machine learning," said >Manoj Nayee, managing director of Boston Limited, in a statement. "We are proud to be a part of the advances made by Intelligent Voice in speech-to-text processing using our custom-made servers."

SpeechTek Covers
for qualified subscribers
Subscribe Now Current Issue Past Issues