May 19, 2021
By Pavel Jiřík in Blog
This year in March, Phonexia released its most accurate speech transcription technology yet: the sixth generation of Phonexia Speech to Text.
Using a novel approach, the latest generation combines together deep neural networks for speech recognition and Phonexia’s cutting-edge deep neural networks for speaker recognition.
Thanks to this unique coupling of voice biometrics and speech recognition deep neural networks, Phonexia Speech to Text’s latest generation offers excellent automatic speaker and channel calibration functionality, achieving exceptional speech transcription accuracy on known as well as unknown voice channels.
At the time of release, Czech was the first and only language supported by the sixth generation. But as we mentioned in the release blog post, other languages were planned to be added and supported very soon.
And so today, we are excited to announce that the Phonexia research and development team has expanded the sixth generation of Phonexia Speech to Text with four more languages!
Four New Speech to Text Languages in the Sixth Generation
Besides the already mentioned Czech language, you can now use Phonexia Speech to Text’s latest generation to transcribe human speech more accurately in the following languages:
Spoken by over 76 million native speakers, French can now be transcribed with up to almost 30% greater accuracy than Phonexia’s previous speech transcription generation allowed.
To be precise, the sixth generation achieves a relative increase of between 23.8% and 29.8% in word accuracy when compared to the previous fifth generation!
This widely spoken language in the Balkans is yet another European language that has made it into the sixth generation.
Compared to the previous fifth generation, it achieves automated transcription with a relative increase in word accuracy of between 11.2% and 19.7%.
Spoken by over 40 million people in Afghanistan and Pakistan, this Middle-Eastern language has also made it onto the list of new languages available in the sixth generation.
Phonexia Speech to Text can now transcribe Pashto with a greater word accuracy, offering a relative improvement of 19.3% over the previous generation.
This widely spoken Arabic language is the fourth language that has made it into the group of the newly released languages for the sixth generation.
Once again, the latest generation of Phonexia Speech to Text transcribes Levantine Arabic with a greater accuracy, outperforming the previous generation by between 16.8% and 23.7% in a relative word accuracy.
Here is a complete list of the new languages and their relative improvements:
More Languages Are Coming
The four languages listed above are only the beginning, as the sixth generation of Phonexia Speech to Text has proven to be an excellent approach to improving the accuracy of automated transcriptions.
Due to the technology’s state-of-the-art arrangement of voice biometrics and speech recognition deep neural networks, an impressive improvement in word accuracy can be achieved for many more languages.
Therefore, in the upcoming months, you can expect other new languages to be released under this currently most accurate generation of Phonexia Speech to Text.