SIGN UP

Enhanced speech recognition model is now available

Enhanced speech recognition model is now available

Voximplant speech recognition for most of the languages is powered by Google Speech API. At the end of February Google introduced new Speech API capabilities including Enhanced speech recognition model for better quality of recognition (for video and phone calls). According to Google, this model offers the 62-64% Word Error Rate (WER) improvement compared to the standard one. 

WER-comparison

It's currently available only for US English, but later enhanced models will support other languages. Voximplant developers can use this model by setting enhanced to true and model to phone_call in ASRParameters:

const asr = VoxEngine.createASR({
	lang: ASRLanguage.ENGLISH_US,
	enhanced: true,
	model: 'phone_call',
	singleUtterance: true
})

Please note that billing for enhanced recognition differs from the standard one as it requires more resources.

 

Tags:ASRspeech recognitionspeech-to-text
B6A24216-9891-45D1-9D1D-E7359CEB8282 Created with sketchtool.

Comments(0)

Add your comment

Please complete this field.

Recommended

Sign up for a free Voximplant developer account or talk to our experts
SIGN UP