Vianix Gets the Year Off to a New SPART
Today, Vianix announced the successor to its automatic speech recognition (ASR) solution, MASC: SPART (Speech Processing for Automatic Recognition Technology).
SPART, which has been in development for more than two years, re-architects MASC from the ground up.
Vianix’s original technology, MASC, focused on providing high-quality files, while also managing limited bandwidth and storage. As its customer base became more sophisticated, though, focus shifted away from data size issues toward attaining a more accurate ASR platform. MASC’s users, particularly in the healthcare industry, began simultaneously demanding higher accuracy ASR and uncompressed quality in playback. This presented a challenge.
According to Veeru Ramaswamy, chief technology officer at Vianix, SPART stands out from its competition because it provides both accurate ASR and high-quality recordings with fully reconstructable voices for playback on the back end. Some firms, Ramaswamy explains, were able to offer high-quality ASR, but were compromising playback readability.
To meet its customers’ twin demands, Vianix re-built the original MASC algorithm to make it suitable for engines to automatically detect high voice quality, while still meeting existing bitrate, processing requirements, and accounting for technological legacy constraints in a given environment (i.e. a standard black phone). To improve analytics and quality, for instance, Vianix introduced noise reduction techniques innately into its algorithm rather than using a standalone filter after the fact.
The results from its new algorithm are, by Vianix’s accounts, dramatic. A press release trumpeting SPART’s debut claims the new system improves voice quality by 10 percent to 15 percent, and improves ASR accuracy by 57 percent to 88 percent. The results, in fact, have been so good for Vianix, that it began finding applications for its technology outside of its medical base.
In voice biometrics, accuracy improvements of more than 1 percent in both true and imposter scores were being found. Vianix also claims improved merit scores in speech analytics of 10 percent to 20 percent.
“[Basically] we’ve been doing this for a long time and in responding to customers, we….completely re-architected our product to address a specific need in healthcare, particularly around dictation/transcription workflows that involve ASR, and did it in away that’s agnostic to the ASR engine provider, and then found that it worked well in some really unrelated markets,” says Bernard Brafman, vice president of sales and marketing at Vianix.”
SPART, which has various tuning capabilities, is now finding applications in public safety, legal, and insurance settings for mobile-based dictation/transcription workflow; call centers; mobile voice services; voice biometrics; and the semiconductor industry, where it’s allowing new devices to become enhanced points of capture—this in addition to improved service of its traditional medical base.