Azure Speech Cognitive Services - Batch Transcription Filler words removal

Question

We are using Azure Cognitive Speech Services v3.1 with Standard tier.

We have a question regarding filler words in the response JSON of Get Transcriptions REST API.

After sending audio/video file to Speech to text Batch Transcription REST API for transcription, We have found the some filler words in the response of Get Transcription REST API.

For Example:

うん。 森さんの机に机の上に山口先生に。 妹本があります。 モーリーソングラスビー使う本を長川先生に引かれました

In the above transcribed content "うん。" is considered as filler word.

Question:

Is there any possibilities to remove the filler words in the response without affecting the actual context of the conversation?

Our Sample Batch Transcription REST API Request JSON:

{     
"contentUrls": [    
"https://example.blob.core.windows.net/transcriptions/example.mp3"
],     
"locale": "ja-JP",     
"displayName": "Transcription Content for File"
}

Answer 1

Hi @Test Admin ,

Thank you for reaching out to Microsoft Q&A.

Regarding your query, it seems that for en-US, Azure Cognitive Speech Services automatically removes filler words. However, for ja-JP, there is no language support documentation for any config parameter that can automatically remove filler words without affecting the actual context of the conversation. Therefore, without any post-processing, it is not feasible to remove filler words from the response without affecting the actual context of the conversation.

I hope you understand. Thank you.