The 2017 Speech Industry Luminaries: Dimitris Vassos

Article Featured Image


Dimitris Vassos

CEO and chief architect of Omilia

For the past 20 years, Dimitris Vassos has been on a mission to destroy the traditional interactive voice response (IVR) framework and reinvent customer self-service. In that pursuit, he recognized early on that the only way to get there was to crack the age-old problem of natural language understanding. He would stop at nothing short of true human-level interaction. This year, he was able to roll out a solution that does just that.

Vassos cofounded Omilia in 2002, and as its CEO and chief architect, he leads a team developing a full stack of state-of-the-art speech recognition, biometrics, natural language understanding, and dialogue manager technologies infused with artificial intelligence for voice, web, and mobile interactions.

Omilia, a 15-year-old company based in Greece, today offers an intelligent dialogue management platform, DiaManT, and an innovative natural language understanding engine, deepNLU, capable of providing a true conversational experience. In addition, Omilia has one of the most accurate on-premises ASR engines (deepASR) in the world, with accuracies in the banking and telecom domains touching human-level performance. Its deepNLU engine delivers greater than 95 percent semantic accuracy in conversational dialogues, and its deepVB biometrics authentication engine can yield a 99.5 percent accurate speaker identification score within just three seconds of any speech sample.

Omilia didn’t enter the North American market until 2015, and within about a year, it was already deploying the first true end-to-end voice portal at one of the continent’s largest banks. Omilia started rolling out the technology to Royal Bank of Canada in October 2016.

Today, Omilia boasts more than 20 call center deployments of true conversational virtual agents in 11 countries, representing 10 languages and serving more than 1 billion customer interactions.

The application goes way beyond the simple intent tagging that typical call steering applications provide. With a true end-to-end conversational interface, there is no predetermined dialogue flow. The interaction is completely unstructured, allowing customers to speak the same way they would speak to other humans. One of the key differentiators of this type of interface is that the technology can maintain context and memory throughout the entire dialogue.

And Vassos knows a lot about dialogues. Prior to founding Omilia, he was a systems engineer at IBM, where he developed and maintained IVR and other voice systems in the United Kingdom. He also served as a senior systems engineer and business development manager at VoiceRite.

SpeechTek Covers
for qualified subscribers
Subscribe Now Current Issue Past Issues