About This Site
Welcome to the Swampscott Open Meetings analysis site, an initiative by town resident Rupert Deese.
Goal: Enhancing Transparency & Accessibility
The primary goal of this project is to make the proceedings of Swampscott's public meetings more accessible and understandable for residents and interested parties. Official town meetings are recorded, but finding specific information within hours of video can be challenging. This site aims to provide useful summaries and analyses derived from meeting transcripts.
How It Works
The process involves several steps, leveraging automated tools and Artificial Intelligence:
- Audio Source: The official video recordings of meetings are downloaded from the Town's YouTube channel.
- Transcription: The audio is processed using an AI transcription service (specifically, OpenAI's Whisper model). This service generates a text transcript and attempts to identify different speakers. While significantly better than basic auto-captions, this process isn't perfect.
- AI Analysis: The generated transcript, along with meeting metadata (like title and date), is then provided to a large language model (Google's Gemini AI). The AI is prompted to generate a structured analysis, typically including sections like:
- Agenda Items Discussed
- Key Decisions Made
- Action Items
- Brief Summary
- Detailed Minutes (based on the transcript)
- Website Presentation: The resulting analysis is formatted and presented on this website, linked alongside the original video recording and the generated transcript.
Important Disclaimers & Limitations
This is an experimental project and not an official record. While the aim is accuracy and usefulness, users must understand the limitations:
- Transcription Errors: AI transcription can misinterpret words, struggle with unclear audio, background noise, or multiple people speaking at once. Names of people, places, and specific terminology are particularly prone to errors. Speaker identification may also be incorrect sometimes.
- AI Analysis Errors: The AI analysis is based *only* on the provided transcript. If the transcript is flawed, the analysis will inherit those flaws. Furthermore, AI can misunderstand context, hallucinate details not present in the source, or misinterpret the nuances of discussion. It may summarize points inaccurately or miss important details.
- Not Official Minutes: The analyses provided here are **not** official meeting minutes. They are AI-generated summaries intended as helpful aids for navigating the meeting content. For official records or decisions, please consult the Town's official sources.
- Verification Recommended: Users should treat the content on this site as a starting point. If a particular detail or statement is important, **always verify it by consulting the original video recording.** Links are provided for this purpose.
Source Code & Contact
This project is provided as an experiment in civic technology and transparency. The source code is available for non-commercial use upon request to Rupert Deese.