Derek Willis

2 months ago | thescoop.org | Derek Willis

Produce a JSON object with the following keys: ‘committee’, which is the name of the committee in the disclaimer that begins with Paid for by but does not include Paid for by, the committee address or the treasurer name. If no committee is present, the value of ‘committee’ should be None. Also add a key called ‘sender’, which is the name of the person, if any, mentioned as the author of the email. If there is no person named, the value is None. Do not include any other text, no yapping.

Jun 21, 2024 | thescoop.org | Derek Willis

It Works! (Mostly)Mostly, this approach worked for me, but sometimes Claude ignored my instructions by giving me introductory text like Here are the JSON objects for the Tangible Gifts extracted from the provided text: before listing the data. I spent time trying to tweak the prompts to get it not to do this, but couldn’t prevent it all the time. In the end, I declared victory and removed the extra text manually. It’s not perfect, but it saved me a LOT of time. How do I know it (mostly) worked?

Jun 19, 2024 | thescoop.org | Derek Willis

If there’s anything that I’ve developed an expertise in in the past 25 years, it’s campaign finance data. And if there’s anything I’ve been really into during the past year, it’s large language models (LLMs). You might think this is a story about how I’ve combined the two. But it’s not, because mostly I don’t, and I’d like to explain why. Because most campaign finance data is structured - in rows and columns - some of the tasks that LLMs are most impressive at doing don’t necessarily apply here.

Jun 2, 2024 | thescoop.org | Derek Willis

I called it my “running with scissors” class, because after I offered to teach a course on using artificial intelligence for journalism, the thought dawned on me that I had very little idea how to actually design and teach such a class. Running with scissors, a terrible image from DALL-EHere’s what I eventually settled on: give students access to various AI services - ChatGPT, Claude, Gemini and other LLMs - and then ask them to apply them to specific journalism-related tasks.

Jan 26, 2024 | thescoop.org | Derek Willis

When it comes to turning official precinct election results into data, there are few states like Mississippi. The Magnolia State isn’t like New York, where there’s no centralized statewide reporting system for precinct results (not yet, at least).