Google Speech API with Cantonese

I am trying to convert some Cantonese audio clip into text with the Google Speech API. The example app on https://cloud.google.com/speech/docs/samples was good (I am using Java) and following the instruction I can get the example converting the sample clip into text. But than I get into trouble converting the Cantonese.

Firstly, I can't get it converting Cantonese clip. It just returned blank result. I did two things to make it return some transcript.

1. Setting the Language Code
   
       RecognitionConfig config =
           RecognitionConfig.newBuilder()
               .setEncoding(AudioEncoding.LINEAR16)
               .setSampleRate(samplingRate).setLanguageCode("yue-Hant-HK")
               .build();
2. Using Audacity and record the clip with MONO channel, and export it as RAW type:
1. File type: Other uncompressed files
2. Header: RAW (header-less)
3. Encoding: Signed 16-bit PCM

And finally I got a response like below:

INFO: Received response: results {

  alternatives {

    transcript: "\350\251\246\345\232\207\345\273\243\346\235\261\350\251\261\350\250\273\345\206\212\346\231\202\351\226\223"

    confidence: 0.8150804

  }

}

It looked like a JSON but it wasn't. It was GRPC's response to get the Chinese transcript we need GRPC client.

I modified the AsynClient a bit and make it return the AsyncRecognizeResponse from the recognize method, and use some code like below:

    AsyncRecognizeResponse result = null;

    try {

      result = client.recognize();

    } catch(Exception e){

    e.printStackTrace();

    }finally {

      client.shutdown();

    }

    result.getResultsCount();

    List rresult = result.getResultsList();

   

    for(SpeechRecognitionResult srr:rresult){

    SpeechRecognitionAlternative alternativesStr = srr.getAlternatives(0);

    String transcriptStr= alternativesStr.getTranscript();

    System.out.println(transcriptStr);

    }

And the transcriptStr contains the correct Chinese characters now.