A friend and I started QCStacks 10 years ago to summarize interesting news articles from various publications around Cincinnati. We would scour local news websites and publications and pick the 5 most interesting links and write a quick summary a few times per week. It was a labor of love, and we never got much traction. We abandoned the effort after 18 months and haven’t picked it up since.
Summarizing Council Meetings
The new iteration of QCStacks will keep its focus on Cincinnati, but instead rely on AI and automation to provide the site’s content. I’m also shifting the focus from a general news roundup to something more specific in summarizing Cincinnati City Council meetings. I’m not particularly drawn to the intricacies of city council, but council’s regular meetings are easily accessible online and provide a number of challenges that I can learn from.
Specifically, the meetings are hosted on Archive.org and contain a complete closed captioning transcript of the meeting. Unfortunately this transcript isn’t formatted for easy consumption, which makes it difficult for a human to read but not an AI model. Council meetings are also boring affairs with little-to-none public interest. Extracting the key highlights and arguments from these meetings and providing readers with a terse summary will provide value for anyone not wanting to sit through the entire meeting, but wants to be informed of what council discussed. The meetings on Archive.org date back to 2015, most of which have a corresponding video and complete closed captioning transcript.
While my initial goal is to summarize future council meetings, there’s potential to go back and summarize everything from the past. This will allow me to further explore other AI application architectures, such as Retrieval Augmented Generation (RAG), to search for specific themes and topics across all meetings from 2015 to today.
Beyond Council Data Sources
While I’m starting with City Council meetings nothing is stopping me from incorporating other open Cincinnati government data into QCStacks. The city of Cincinnati publishes a large amount of government data ranging from police reports to economic initiatives on Open Data Cincinnati available for public consumption. Most of this data is unintelligible unless you have a data science background or have a knack for finding interesting trends.
But with the help of large language models, specifically ones similar to the Code Interpreter from ChatGPT, we can automate the work to find interesting insights.
I’d like QCStacks to become a library of accessible knowledge for Cincinnati residents to stay informed of their community. Think of it as a continuation of EveryBlock, the now-defunct hyperlocal news platform, but enhanced with artificial intelligence. And while EveryBlock failed as a commercial product, I’m not looking to make a buck from QCStacks. My goals for this project are to increase my learning while also providing value to Cincinnati by making local government meetings and data accessible to everyone.
I wanted to write this summary of what QCStacks is for both myself and anyone else following along to have a bit more context. I’m looking forward to the new AI enhanced iteration of QCStacks and hopeful that we’ll get a bit more traction this time around.