DNN Speech Recognition


Our Automatic Speech Recognition engine leverages the most advanced forms of Deep Learning, achieving unprecedented accuracy in recognition that routinely reaches human-level performance.

deepASR® Languages

Today, our engine excels in recognizing 17 languages including English (US, Canada, UK, & South Africa), Spanish, Russian, Polish, Kazakh, Ukrainian and Greek.

Thanks to Omilia’s proprietary method of training and tuning, deepASR® is able to achieve Word Error Rates of less than half of legacy incumbent providers.

For all primary languages Omilia offers adapted acoustic and  language models that cover the accent and dialectic variations within the country.

English - US

English - UK

English - Canada

English - South Africa

English - Caribbean

French - Canada

Spanish - Spain

Spanish - Latin America

Russian - Russia

Russian - Ukraine

Russian - Belarus

Russian - Kazakhstan

Polish - Poland

Kazakh - Kazakhstan

Ukranian - Ukraine

Mixed Ukranian-Russian - Ukraine

Greek - Greece

And more...

Don’t see your language here?  We will develop an adapted acoustic and language model for your language in less than 2 months.

Achieving Human-like Results

Omilia’s deepASR® was developed to offer our customers a complete solution for natural language understanding while also focusing on the return on investment of the project. Since the legacy ASR providers offered sub-par speech-to-text solutions at ridiculously high prices, developing a proprietary ASR engine was the key to bringing bottom line value to our clients operations.

Why deepASR® succeeds where others fail?

Your customers do not speak one single language — in reality your customers have a very wide range of accents and ways of expressing themselves. In todays globalized economy there is no “one size fits all” for any language model. In the past strong accents, slang and ethnic vocabulary make companies nervous about new speech technologies. This reservation towards speech technologies stems from over-promised and under-delivered solutions from our competitors, that just didn’t quite work outside their lab.

In many cases the sound quality reaching the call center can be very poor due to many reasons — because most recognition engines are trained in a laboratory to understand perfect quality sound, they inevitably fail in the real world where sound quality is usually sub-par. Omilia has solved this problem by training our recognition models with real world call center audio to optimize the language and acoustic models of our ASR engine.  With this personalized approach to speech recognition Omilia reached unprecedented accuracy in speech to text transcription.

Arrange a demonstration

Our proven Omni-Channel technology is aimed at:

  • Large Corporations (200+ agents / 4+ million calls per year)
  • Integrators & Contact Center Service Providers

If you represent a relevant business and would like to arrange a demonstration of our technology and learn how it can transform your customer care, fill out our form and we will get in contact with you to get the ball rolling.

  • This field is for validation purposes and should be left unchanged.