Speechmatics raises bar with new speech recognition technology
Cambridge-based Speechmatics has launched a new real-time virtual appliance designed to improve automatic speech recognition across the industry.
The software has been created with the intention of building the best speech technology for any application, anywhere, in any language.
The company says that whilst continuous transcription technology has come a long way, current offerings aren’t versatile enough to perform effectively across a range of criteria – whether this is higher levels of accuracy, automation or an inability to display an ongoing, current conversation with low latency.
Due to these failings, current technologies are useless in situations where real-time awareness is required, says the company.
With the new Real-Time Virtual Appliance, users can plug Speechmatics’ new software directly into an audio stream and get instantaneous real-time transcription.
The technology has been built from the ground up to offer the same capabilities in real-time as it already does in the batch cloud transcription service, something currently unrivalled in the market.
To further increase accuracy the appliance adjusts words as the context within a sentence becomes clearer. Users can adjust the balance between delay and accuracy and also use Speechmatics dynamic transcript parameters to achieve the best accuracy possible to meet their specific needs.
This flexibility allows Speechmatics’ technology to be used in everything from transcribing speeches to bulletins, improving on the current captioning standards whilst retaining the familiar, continuous stream style, the company claims.