“VoiceBase correctly identified 92% of instances of the selected names compared with 85% in your current phonetic search.” Traditionally phonetic speech engines have stayed around despite the significantly lower transcription accuracy because of their ability to spot out of vocabulary words better than dictionary based methods. With VoiceBase’s breakthrough feature, Custom Vocabulary, the legacy phonetic based engines can finally be retired and completely replaced by state of the art deep learning speech engines.
“VoiceBase’s contribution to Veritone’s CMP is central to our machine learning capabilities, allowing us to analyze recorded audio with unprecedented granularity,” said Chad Steelberg, Chairman and CEO of Veritone. “Their highly accurate transcription quickly recognizes specific words being spoken in a recording that can transform audio into actionable intelligence. Finding relationships and trends between words provides insight that was previously extremely difficult to extract, giving our customers an edge over their competitors,” added Steelberg.Learn More
Rapid Search Results
VoiceBase speech-to-text uses deep
learning, neural network technology to
automatically transcribe and search audio
and video files. Unlike phonetic systems
search results can be delivered instantly,
while using very little computing power.
Instantly Detect Out of Vocabulary Terms
VoiceBase’s unique custom vocabulary feature
was developed to eliminate a limitation with LVCSR. Now users can instantly add out of vocabulary words to optimize speech-to-text results and correctly spot names, brands, products and other terms which are not in the trained speech-to-text engine.
Instantly mine broadcast media for
name and product mentions
Make minimally informed decisions
instantly search for additional terms