The accuracy of Japanese Kanji conversion is less compared to V1.(Testing in Long model)
The use of the Chirp model increases the accuracy for the Japanese language but when using StreamingRecognize, Chirp does not support the use-case.[1]
Expected behavior
Extend the support for streaming recognize for "chirp_2" model.
Actual behavior
Chirp does not support streaming recognize API
Current customer workaround
The customer tried using model adaptation to increase the accuracy but the model adaptation is not taking effect for Japanese Kanji
Description
Issue summary
The accuracy of Japanese Kanji conversion is less compared to V1.(Testing in Long model)
The use of the Chirp model increases the accuracy for the Japanese language but when using StreamingRecognize, Chirp does not support the use-case.[1]
Expected behavior
Extend the support for streaming recognize for "chirp_2" model.
Actual behavior
Chirp does not support streaming recognize API
Current customer workaround
The customer tried using model adaptation to increase the accuracy but the model adaptation is not taking effect for Japanese Kanji
Documentation consulted
[1]https://cloud.google.com/speech-to-text/v2/docs/chirp-model#available_api_methods