Summarising Teams Meeting Transcripts using Open AI and Python

Time-consuming meeting recordings and extracting key points can be a thing of the past. This Python-based solution uses Azure Open AI to automate the summarisation process for Teams meeting transcripts. The tool efficiently breaks down meetings into five key areas. These areas include Overview, Concerns, and Current Model. It also covers Questions/Thoughts to Address and To-Do Lists and Outstanding Tasks.

Tools like Microsoft Copilot for Teams offer AI-driven capabilities. However, they may not always suit specific customisation needs. They might also not fit within budget constraints. To address this, I developed a Python-based solution leveraging Azure OpenAI for transcript summarisation.

Key Features

This tool summarises meeting transcripts into five core areas:

• Overview
• Concerns
• Current Model
• Questions/Thoughts to Address
• To-Do Lists and Outstanding Tasks

How It Works

The process begins by reading the Teams transcript file. The script divides the document into smaller “chunks” for efficient AI processing. Each chunk is processed by an Azure Open AI model, generating a summary for that specific section. Finally, the script consolidates these individual summaries into a final, organised summary of the entire transcript. This approach provides accurate results without the need for complex vector databases.

Lets look at this step-by-step:

1. Transcript Parsing: The tool reads a Teams transcript file
2. Chunking: Splits the transcript into smaller sections, ensuring manageable input for the AI model
3. Summarisation: Each chunk is processed using Azure OpenAI, which generates summaries for each segment
4. Consolidation: All chunk summaries are combined into a structured final summary, categorised into the five key areas.

Benefits

1. Rapid summarisation: The tool can summarise a 1-2 hour meeting in approximately 60 seconds
2. Clear overview: Provides a concise summary of each recording, highlighting areas of interest
3. Easy navigation: Timestamps for each user and topic allow quick access to relevant sections
4. Time-saving: Significantly reduces the time needed to extract key information from meetings
5. Enhanced Knowledge Sharing: With clear, categorised summaries, it’s easier to revisit specific topics or discussions. Timestamps provide direct references for detailed follow-ups.
6. Full customisation: This Python-based tool can be adapted to unique organisational needs, ensuring relevance across various use cases. I have provided five core areas that the Teams transcript can be summarised into, you can change/add/modify this within the prompt to your requirements

Why Use This Over Microsoft Copilot for Teams?

While Microsoft Copilot for Teams offers powerful features, this custom solution provides several advantages:

• Integration potential: Easily integrate with other custom tools and workflows.
• Customisation: Tailor the summary format and categories to your specific needs.
• Flexibility: Copilot performs best in specific languages. However, this tool can be adapted for any language supported by Azure Open AI.
• Cost-effective: Potentially more economical for organisations not subscribed to Microsoft 365 Copilot.

Example Summary

Below is an example summary from the Python tool (Please note: Data is anonymised with both names and content)

Overview:

- Emma (3:12): Discussed updates on cloud storage policies and permissions.
- Liam (10:25): Outlined strategies for improving deployment automation.

Concerns:

- Olivia (15:47): Raised issues about inconsistent deployment timings across regions.

Current Model:

- Noah (3:12): Focus on containerization efforts for microservices.
- Sophia (11:03): Shared progress on migrating legacy databases to the cloud.

Questions/Thoughts to Address:

- Ava (15:20): Is there a standard format for API documentation?
- Mason (16:45): Can monitoring dashboards be shared across teams?

To-Do Lists:
For Isabella:
- Coordinate with Ethan on updating backup retention policies.

Outstanding Tasks:
- Network Configuration Testing: Validate the updated firewall rules.
- Documentation Review: Ensure compliance with new security standards.

• Notice the use of timestamps? Very useful to go back to the meeting recording to review further if required.

GitHub Repository

The Python solution is stored in this repository including setting up and how to run.

Also see the prompts that have already been setup:

• Overall Summary
  • Purpose: Generates a general summary of the entire Teams transcript
• User-Specific Summary
  • Purpose: Creates a summary with references to each participant and their contribution timestamps
• Comprehensive User Summary with Q&A
  • Purpose: Produces a detailed summary including user references, timestamps, and addressed questions

Feel free to update to suit your own requirements

Wrapping up

Automating the summarisation of Teams meeting transcripts with Python and Azure OpenAI gives you a powerful tool. This tool enhances efficient knowledge management. This solution offers cost-effective alternatives. It provides customisable and private options to out-of-the-box tools like Microsoft Copilot for Teams. These features empower organisations to focus on what matters most.

Ready to save time and streamline your meeting notes? Try implementing this approach today!