Google Speech API metadata doesn't effect the results or conversion method

1
votes

I am using the following funtion parameters to convert speech to text but the extra parameters i set for metadata parameters doesn't have effect on the conversion result with or without parameters.

function transcribe_async_gcs($bucketName, $objectName, $languageCode = 'en-US', $options = ['enableAutomaticPunctuation'=> true,'enableWordTimeOffsets'=> false,'encoding'=> 'FLAC','sampleRateHertz' => 44100, "metadata"=> [ "interactionType"=> "PHONE_CALL", "industryNaicsCodeOfAudio"=> 621210, "microphoneDistance"=> "NEARFIELD", "originalMediaType"=> "AUDIO", "recordingDeviceType"=> "PHONE_LINE", "originalMimeType"=> "audio/mp3", "audioTopic"=> "appointment"],'useEnhanced' => true])

1
speech-recognitionspeech-to-textgoogle-speech-apigoogle-cloud-speech

1 Answers

0
votes

In order to increase speech recognition accuracy, apart from metadata [1], you can provide phrase hints with your configuration [2]. This may increase the accuracy if you know the context of the speech. There is also an option to use enhanced speech recognition model for better quality results [3], in which you need to opt-in data logging and be able to use special set of machine learning models. These are all optional fields and regarding to the audio content may or may not have an effect on the results.