Welcome to part 19 of the tutorial series on AWS Audio Analysis. In this tutorial, the helper lambda function is updated to separate speaker transcripts from Amazon Transcribe output and store back into the S3 bucket.

GitHub code: https://github.com/srcecde/aws-tutorial-code/blob/master/lambda/lambda_transcribe_speaker_seperation.py

---
Support my work:
---
Patreon: https://www.patreon.com/srcecde
PayPal: https://paypal.me/srcecde
Paytm | Gpay: 9023197426 

---
Series Tutorial
---
Architecture diagram: https://youtu.be/y_j-r8x1FtI
Part 1: https://youtu.be/Jz8NuapOI3I
Part 2: https://youtu.be/S06cqct732g
Part 3: https://youtu.be/tgOYpPr2cic
Part 4: https://youtu.be/E1NGS4B2mr0
Part 5: https://youtu.be/lTJR0NUKebw
Part 6: https://youtu.be/s7rtX5EQSVg
Part 7: https://youtu.be/uqOqBXcvPgk
Part 8: https://youtu.be/tkxany-M3I8
Part 9: https://youtu.be/lsqAmC41C5E
Part 10: https://youtu.be/Pe7JoTyAkzo
Part 11: https://youtu.be/dHwY-K-1y9I
Part 12: https://youtu.be/pSb0nwcyFBs
Part 13: https://youtu.be/RFrJh0uAW68
Part 14: https://youtu.be/kwrDE7bXLKU
Part 15: https://youtu.be/u6LT3I5XYw8
Part 16: https://youtu.be/G2XuT8UDvl4
Part 17: https://youtu.be/48y7PbZ-9sM
Part 18: https://youtu.be/Fd5_frV5TVo
Part 19: https://youtu.be/4qGmRgO2WNE
Part 20: https://youtu.be/Nnblz5qgMGA
Part 21: https://youtu.be/nLJmTfaMdHg
Part 22: https://youtu.be/RPhOuvJGdPo
Part 23: https://youtu.be/3x-zdNZYFjo
Part 24: https://youtu.be/F2JUkiuhRTI
---
Another channel:
---
My Gaming Channel: https://www.youtube.com/channel/UC8QMD-3yuE-_0h-ysWrqZhw 

---
Connect with me
---
Twitter: https://twitter.com/srcecde
GitHub: https://github.com/srcecde
Facebook: https://www.facebook.com/SrceCde/
Instagram: https://www.instagram.com/srcecde/
LinkedIn: https://www.linkedin.com/company/srcecde/
Reddit: https://www.reddit.com/r/srcecde

What's up guys and welcome to part 19 of tutorial series on AWS audio analysis.

In the previous tutorial we have created the new resources that were marked in red as per our architecture diagram as you can see on my screen.

And also we had updated and set the correct triggers.

Hence our flow is all set as per our new architecture diagram or as per our new use case.

So now in this tutorial, we are going to update our lambda function to derive separate transcript of individual speakers after speaker identification or speaker diarization So before we start let me first give you an idea that why we need this helper lambda function and why we are going to update it.

we look at the architecture diagram, then Amazon transcribe will give us an output Within an s3 bucket something like this so I will wrap it here, So this is the overall transcribe of an audio.

And after that here you can see the speaker label.

And what he or she is speaking then we have speaker 1 right and and this is how the output of transcribe is so here we require an individual transcript for speaker 0 and speaker 1.

So that is agent and customer in our case.

So as you can see it's quite messy to even look at it.

So let me try to spot some speaker level transcription for you.

So, so for example, we are considering the speaker and the school 1.

So maybe not that I am bit confused where we are.

So what, what you can see here is it's just the start time and end time that is being referencing

to an individual speaker right so now if I look for this time stamp so here here you can see within items right So the start time is

and we have a content over here that this word has been spoken within this given time interval.

So and we don't need this messy output to be fed into a comprehend for sentiment analysis, right?

So that's where We require the outcome like this.

So as you can see, this is the speaker 0 transcription.

So for example, agent and customer separate transcript.

So that is the reason we are here to update our or write some code within the lambda function that is our helper lambda function to get such kind of output.

So when I say such kind it's transcription of an individual speaker bundle up separately.

Right, so we will move back to our AWS management console.

So just to remind you that we are updating this AWS Lambda function that we have created in the previous tutorial.

So let's navigate to AWS management console and jump to Lambda management console.

So this is the lambda function that is AWS Audio Analysis Helper that we have created in the previous tutorial.

So here I am not going to write a code from scratch, but I have already written the code so I'm going to simply copy and paste.

So this is the lambda code that we are going to use in order to generate individual transcript for an individual speaker.

So I'm going to copy this go back to our lambda function and I'm going to paste this and I will save this.

I will take you through this code at high level.

So here we are using 2 libraries that is JSON and boto3.

11 I think you will be familiar with these lines.

So this is nothing but the s3 bucket trigger and based on that trigger we are reading the file and loading it into this transcript on the school result.

what we are loading in transcript underscore result

is this file that is transcribed output right.

So we are loading this data or reading this data from an S3 bucket on trigger.

Now post this starting from line 13 to line number 38.

logic to get the file something like this.

I mean to get the transcript of an individual speaker.

So I'm doing nothing but an JSON file or the dictionary manipulation over here in order to come up with these 2 separate transcripts.

So this logic even works for if you have for example more than 2 speakers.

So you can use this code in your use case too in case if you have more than 2 speakers.

And finally post retrieving the transcript of an individual speaker, I'm uploading it to the S3 bucket.

Now this S3 bucket refers to this 1 that is Amazon S3 over here within our architecture diagram again we have created this S3 bucket in the previous tutorial so let me go ahead and show you

So What this lambda function will do post?

Separating the transcript of an individual speaker.

It's going to upload the json file for individual speaker.

So so it will bundle up all the speakers into a single json file labeling that speaker underscore 1 and its transcripts speaker underscore 2 and its transcripts.

So that's how it will create a JSON file or the JSON data and that's what we are uploading it to this S3 bucket now this S3 bucket will trigger our another lambda function that is responsible for sentiment analysis so that is something we are going to cover in the next tutorial, right?

So and This helper lambda function again.

It will help us in retrieving the transcript of an individual speaker and that's all for now And in the next tutorial we are going to update or modify the lambda function.

As I say, that is responsible for sentiment analysis as per our requirement.

And as usual, if you want me to do tutorial on any use case or service, then please leave them below and I will try my best to come up with a tutorial as soon as possible.

And if you have any queries or comments, then again, please leave them below.

And don't forget to like and subscribe to my channel and see you next time.


Speaker 0: What's up guys and welcome to part 19 of tutorial series on AWS audio analysis. In the previous tutorial we have created the new resources that were marked in red as per our architecture diagram as you can see on my screen. And also we had updated and set the correct triggers. Hence our flow is all set as per our new architecture diagram or as per our new use case. So now in this tutorial, we are going to update our lambda function to derive separate transcript of individual speakers after speaker identification or speaker diarization So before we start let me first give you an idea that why we need this helper lambda function and why we are going to update it.

So let's have a look.

Speaker 1: So this is the transcribe outcome. So if

Speaker 0: we look at the architecture diagram, then Amazon transcribe will give us an output Within an s3 bucket something like this so I will wrap it here, So this is the overall transcribe of an audio. Right. And after that here you can see the speaker label. Right. And then we have speaker 0.

That is and time and start time. Right. And what he or she is speaking then we have speaker 1 right and and this is how the output of transcribe is so here we require an individual transcript for speaker 0 and speaker 1. So that is agent and customer in our case.

Speaker 1: So as you can see it's quite messy to even look at it. So let me try to spot some speaker level transcription for you. So, so for example, we are considering the speaker and the school 1. So maybe not that I am bit confused where we are. So, so let me pick up this start time.

That is

Speaker 0: 0.84. So what, what you can see here is it's just the start time and end time that is being referencing

Speaker 1: to an individual speaker right so now if I look for this time stamp so here here you can see within items right So the start time is

Speaker 0: 0.84

Speaker 1: and we have a content over here that this word has been spoken within this given time interval. So this is something like this, right? So and we don't need this messy output to be fed into a comprehend for sentiment analysis, right? So that's where We require the outcome like this. So as you can see, this is the speaker 0 transcription.

And this is the speaker 1 transcription. So for example, agent and customer separate transcript. So that is the reason we are here to update our or write some code within the lambda function that is our helper lambda function to get such kind of output. So when I say such kind it's transcription of an individual speaker bundle up separately. Right, so we will move back to our AWS management console.

So just to remind you that we are updating this AWS Lambda function that we have created in the previous tutorial. So let's navigate to AWS management console and jump to Lambda management console. So here we are. So this is the lambda function that is AWS Audio Analysis Helper that we have created in the previous tutorial. So here I am not going to write a code from scratch, but I have already written the code so I'm going to simply copy and paste.

So this is the lambda code that we are going to use in order to generate individual transcript for an individual speaker. So I'm going to copy this go back to our lambda function and I'm going to paste this and I will save this. I will take you through this code at high level. So here we are using 2 libraries that is JSON and boto3. Now from line number 5 to line number

Speaker 0: 11 I think you will be familiar with these lines. So this is nothing but the s3 bucket trigger and based on that trigger we are reading the file and loading it into this transcript on the school result. So

Speaker 1: what we are loading in transcript underscore result

Speaker 0: is this file that is transcribed output right. So we are loading this data or reading this data from an S3 bucket on trigger. Now post this starting from line 13 to line number 38.

Speaker 1: So that is this chunk of code.

Speaker 0: So here I have wrote a

Speaker 1: logic to get the file something like this. I mean to get the transcript of an individual speaker. So I'm doing nothing but an JSON file or the dictionary manipulation over here in order to come up with these 2 separate transcripts.

Speaker 0: That is speaker 0 and speaker 1.

Speaker 1: So this logic even works for if you have for example more than 2 speakers. It is not limited to only 2 speakers. So you can use this code in your use case too in case if you have more than 2 speakers. Right. And finally post retrieving the transcript of an individual speaker, I'm uploading it to the S3 bucket.

Now this S3 bucket refers to this 1 that is Amazon S3 over here within our architecture diagram again we have created this S3 bucket in the previous tutorial so let me go ahead and show you

Speaker 0: So here we are and

Speaker 1: this is our s3 bucket, right? So What this lambda function will do post? Separating the transcript of an individual speaker. It's going to upload the json file for individual speaker. So so it will bundle up all the speakers into a single json file labeling that speaker underscore 1 and its transcripts speaker underscore 2 and its transcripts.

So that's how it will create a JSON file or the JSON data and that's what we are uploading it to this S3 bucket now this S3 bucket will trigger our another lambda function that is responsible for sentiment analysis so that is something we are going to cover in the next tutorial, right? So and This helper lambda function again. It will help us in retrieving the transcript of an individual speaker and that's all for now And in the next tutorial we are going to update or modify the lambda function. As I say, that is responsible for sentiment analysis as per our requirement. And as usual, if you want me to do tutorial on any use case or service, then please leave them below and I will try my best to come up with a tutorial as soon as possible.

And if you have any queries or comments, then again, please leave them below. And don't forget to like and subscribe to my channel and see you next time.

See all Srce Cde transcripts on Youtube

Separating Speaker transcript - AWS Audio Analysis p19