I am testing AWS Transcribe for my PoC project.
When using "speaker partitioning" and "custom vocabulary" at the same time, the results.items
section does not contain speaker_label, only for the item that matches the custom vocabulary:
{
"jobName": "transcription-job",
"accountId": "MY_ACCOUNT_ID",
"status": "COMPLETED",
"results": {
"transcripts": [ ... ],
"speaker_labels": [ ... ],
"items": [
{
"type": "pronunciation",
"alternatives": [
{ "confidence": "0.8", "content": "WORD_INCLUDED_IN_VOCABULARY" }
],
"start_time": "1.23",
"end_time": "1.419" // speaker_label is not present only for vocabulary matches
},
{
"type": "pronunciation",
"alternatives": [
{ "confidence": "0.8", "content": "WORD_NOT_INCLUDED_IN_VOCABULARY" }
],
"start_time": "1.540",
"end_time": "2.210",
"speaker_label": "spk_0" // speaker_label is present, as expected
}
]
}
}
Is this an expected behavior, or am I missing something? I am using @aws-sdk/client-transcribe@3.523.0
and ja-JP
for the language settings. The request is sent like this:
await transcribeClient.send(
new StartTranscriptionJobCommand({
TranscriptionJobName: jobName,
Media: { MediaFileUri: S3_URI },
OutputBucketName: bucketName,
OutputKey: "my-output-files/",
LanguageCode: "ja-JP",
Settings: {
ShowSpeakerLabels: true,
MaxSpeakerLabels: 10,
VocabularyName: awsVocabularyName,
},
})
);