/call_center_ai

Primary LanguagePythonMIT LicenseMIT

Contact Center AI Solution with GPT

This solution demonstrates the application of Microsoft AI and GPT to help solve challenges common to contact center scenarios

Author: Jixin Jia (Gin)
Created: 2023/03/01
Version: 1.4


Key Features

  1. Transcribe conversations in real-time
  2. Simultaneously translate into multiple languages
  3. Identify speaker (diarization)
  4. Analyze speech sentiment in live
  5. Summarize conversation and get a glance at important topics

About

  • This WSGI demo app runs on single instance of Gunicorn for production workload. It is recommended to build it on a docker orchestration service if you target a more scalable workload.

  • All transaction runs within browser (client end). There are no data transmitted to or stored at the backend server in this demo.

  • To run the demo following Azure services must be provisioned and provided at run time:

Azure resource Purpose Required information
Cognitive Service Speech Services For live speech-to-text, speaker diarization and translation key, endpoint, spoken language, translate target langage
Cognitive Service Language Services For real-time sentiment scoring and key named entity recognition key, endpoint
Azure OpenAI Service For generating call summary and assign a topic to the transcribed call key, endpoint, deployed gpt model name

See it in action

Animated Demo


Setup Guide

0. Install

Install dependencies. This solution has been tested on Python 3.7~3.10

pip install -r requirements.txt

1. Run the app

For Unix:

./run.sh

For Windows:

waitress-serve --listen=0.0.0.0:80 wsgi:app

2. Open your favorite browser

Type following in address bar:
localhost:80


3. Add keys into Setup page

Setup screenshot


4. Press mic button and start talking to it

Setup screenshot

Enjoy the demo !


Credit

Special thanks to Nobu Tanahashi for his ingenious Conversation Transcription source code and idea used in this demo.