The 2017 Speech Industry Luminaries: Yishay Carmiel
THE AI CONVERSATION STARTER
president of Spoken Labs at Spoken Communications
Conversational artificial intelligence and deep learning have led to dramatic improvements in the performance of speech recognition, machine translation, and language understanding, among other technologies. But significant challenges remain. One of the biggest comes when trying to implement a cutting-edge algorithm using an open-source, deep learning framework that can operate at large scale. It’s a challenge that Yishay Carmiel, president of Spoken Communications’ Spoken Labs in Seattle, has accepted willingly, and the results so far have been impressive.
At Spoken, Carmiel leads the team working to advance deep learning and machine learning technologies for speech recognition, speech profiling, natural language processing, computer vision, and large-scale data analysis. That team recently concluded a research project with Johns Hopkins University’s Center for Language and Speech Processing. The group investigated an end-to-end text-independent speaker verification system. Carmiel and the team created a revolutionary speaker-embedding system based on innovative deep learning models. Given a large number of training speakers, the system outperformed other state-of-the-art systems in error and miss rates by more than 30 percent.
Carmiel’s efforts have also led to a system that can analyze 1 million hours per day of conversational data and takes advantage of ultra-large-scale data and recent breakthroughs in neural-based speech representation.
Among his goals, Carmiel is working to see artificial intelligence take greater shape in the call center, where it has great potential to streamline automation, improve the customer experience, and respond to the 24/7 demands of modern customer support. The challenge, he says, lies with implementing AI into agents’ day-to-day applications and understanding what other things can be extracted from speech, such as emotion detection.
Another of his stated goals is real analysis of conversations in real time across thousands of simultaneous voice interactions.
Carmiel, though, is up for the task. He has 15 years of experience as an algorithm scientist and technology leader. He is the founder and former CEO of Santiaga Technologies, an embedded speech recognition provider, and former CEO of S-Infinity, a provider of deep learning for conversational AI. He has been with Spoken since August 2015.