Blog - Company

May 4, 2022 | Read time 2 min

BOLD Awards 2022: Speechmatics Takes Gold for BOLDest AI

Speechmatics’ Machine Learning team has been awarded BOLDest AI at the BOLD Awards for their innovative and inclusive speech recognition.
Speechmatics Team Meet the Team

BOLD Awards 2022: Speechmatics Takes Gold for BOLDest AI

There’s nothing quite like industry recognition to know you’re on the pathway to something extraordinary. On Friday 22nd May 2022 in Venice, Speechmatics was awarded the prestigious BOLD Award for BOLDest AI, thanks to their incredible work in speech recognition and artificial intelligence.

Making a Positive Impact

Celebrating their third year, The BOLD Awards recognize the best companies, projects, and individuals in technology. They look out for – and reward – those leading the way in the digital industry, paying particular attention to those who make a positive impact on the world around them. Something Speechmatics is constantly striving to do.

Represented by John Hughes and Benedetta Cevoli, (Accuracy Team Lead and Data Science Engineer respectively), Speechmatics flew to Venice for the awards ceremony. With characteristic humility, they downplayed their expectations of winning, despite their pride in the amazing work they’ve achieved. “The competition is so strong,” Benedetta said, before the event, “just being nominated is a tremendous achievement.”

Solving Real Problems

But win they did, taking home the top prize ahead of smart work by Wesii, Waylay NV, Saipem and Lyriko. The category itself, BOLDest AI, was categorised as being, “open to any Artificial Intelligence (AI) standalone project or as part of a larger project.” The success of the project was determined by “how the development of AI led to improved performance or its ability to solve real problems.”

The “real problem” Speechmatics wanted to overcome was the bottleneck caused by the lack of labeled data in ASR. The solution? Developing our own family of self-supervised speech algorithms and applying them to speech recognition technology. This large-scale network (known as Autonomous Speech Recognition) leads the industry as the first commercially available speech recognition engine built on self-supervised acoustic models. The team firmly believe that self-supervised acoustic models will become the go-to for ASR and thus, Speechmatics are at the forefront of a bold, new way of doing things.

Seeing Real Results

The results from the testing part of the Autonomous Speech Recognition project speak for themselves. Inclusivity was always a major goal for this project and over the past year, Speechmatics has seen uplifts in accuracy across a diverse range of voices. We’ve also seen huge improvements in environments – such as those with noisy backgrounds – typically poorly served by ASR.

Speechmatics’ Autonomous Speech Recognition” has been a labor of love for all sectors of the business, and seeing it rewarded in this way has been a huge boost for all. As John said before the flight home, "we had no expectations going in and were completely “BOLD” over when we won the award".