How is Audio Identification transcription text created from JSON?

Question

Dislaimer:  I'm not a CS major :(   
     
My son records lectures (1-5 people) in college.  With AWS Transcribe, I've upload an audio file (in buckets), and transcribed it.
     
The transcription ("asrOutput.json") is a file he can open in Notepad, copy to Word, and format (id speakers, adding paragraph breaks, etc.).    
*However*, I found that "Speaker Partitioning" breaks the long body of text into various speakers, making the output/reformatting much easier for him (http://threadgill.com/TranscriptionPreview.jpg).    
But how do you have AWS built-in Lambda do that?   
This question was asked in 2018 (https://github.com/trhr/aws-transcribe-transcript), but the answer's step-by-step instructions weren't applicable in today's UI for Amazon S3 > Buckets, or Amazon Transcribe > Transcription.    
Per the answer, S3 has a Lambda function that triggers when uploading arsOutput.json into an "input/" folder, creating a Speaker Partioned formatted file in the "output/" folder.    
Per the github instructions, I created an S3 bucket with two folders:  input/ and output/ , but was unable to properly configure the Lambda function.    
Any help would be much appreciated!

SDillon · Answer

It's easier to follow the [Regular Directions](https://github.com/faangbait/aws-transcribe-transcript#regular-directions) in that Github post (not the Lamba directions) and:
* Download the .py file from the repo
* Ctrl click the folder in Finder to open a Terminal window where you saved the JSON transcript file
* Run `python ./transcript.py asrOutput.json` — I had to run python3 instead of python
* The transcript file is saved as a .txt file in the same folder with the speakers separated

Thanks faangbait!

Riku_Kobayashi · Answer

I believe you can set it up using the method in this document.  
https://docs.aws.amazon.com/AmazonS3/latest/userguide/enable-event-notifications.html

When configuring, set "input/" in "Prefix." and check "All object create events" in "Event types".  
![event](/media/postImages/original/IM81b6y_23RiK5ct-4KWmI4A)

Then scroll to the bottom of the screen and under "Destination" you can select Lambda.  
![lambda](/media/postImages/original/IMGIwvMGA3Q3-G5J_1lcQINQ)

How is Audio Identification transcription text created from JSON?

Add your answer

Relevant content