IssueTracker

What you would like to accomplish:
- I’d like to have an option to distinguish certain types of numbers in ja-JP in Speech-to-Text.
- For example, the current Speech-to-Text model ”speech:recognize” in ja-JP can’t distinguish ”きゅうじゅうきゅう” (“kyu-ju-kyu” which means ninety-nine) from “きゅうきゅう” (“kyu-kyu” which means nine-nine) as the model recognizes both as “99”. They sound different, but the transcribed character is the same.
- Although this is a ja-JP specific issue, it would be desirable if there were an option to distinguish these two, such as transcribing them as “九十九” (“kyu-ju-kyu”) and “九九” (“kyu-kyu”) respectively.

How this might work:
- In fact, these characters are treated differently in some cases in Japanese, and it’d be beneficial for certain use cases where users need to distinguish these numbers.

If applicable, reasons why alternative solutions are not sufficient:
- There is no model that has the ability to do so at the moment.

Other information (details about model and method):
– Model: `command_and_search`
- Version: `v1p1beta1`
- Request method: https://speech.googleapis.com/v1p1beta1/speech:recognize