Inspiration

I'm doing several demos to promote cognitive services and ML.NET. I wanted to make a demo that is more relatable to big companies.

I heard some companies have problem in figuring out how is their customer support doing and what is the general gist of the conversations. They have thousands of hours of recorded audio.

What it does

It converts speech to text, sentiment, key phrase and named entities analysis.

How I built it

I used .NET Core, Speech and Text API from MS Cognitive Services.

Challenges I ran into

It was easy bearly a challenges.

Accomplishments that I'm proud of

It's an awesome demo to show off on user groups!

What I learned

Speech API has a Nuget package that already handles a lot of things for us, like getting audio from microphone, and audio files and sending them to the API to translate them to text.

Text API Nuget package wasn't as nice and I just used raw HTTP requests instead.

What's next for Demo for Customer Service PoC

I want to get a huge dataset, process it and store the metadata to a DB. Then I want to use ML.NET to cluster conversations for easier analysis. Also, I would like to give operators suggestions on what to say based on the conversation they had with the customer so far.

Built With

Share this project:

Updates