A couple weeks ago all 6 weeks of the Depp V heard transcripts were unsealed. This includes all the side bars. However, they’re all image PDFs. We all know that I followed that case last year, when it was televised. I really don’t need to read the entire transcripts over again, just all the side bars that weren’t streamed. It’s easy to fix image PDFs with screen readers that hve built-in OCR programs. In fact, that’s what I use to fix most of them now. However, it does present a structuring issue. If I could put something in the text files to tel me when the side bars were happening, I’d be able to find them easier. Wait, if I could make them into web interfaces, and read them in the browser, I could read them easier. If I could put HTML headings when a side bar begins, I could really start to have a good document structure happening. That would take me days. Probably weeks, maybe longer, depending. If I could find someone to do this for me. Nah, too expensive. If I could find something to do this for me. Hey, what about chat GPT? I’ve never really used GPT in my life, except for to play around with it for like a half hour or so. This is how the process will most likely go. My OCR program will turn the image PDF into cleartext, so the screen reader can read it. I’ll take the text out of the virtualized results buffer and place it into a text file so that I have kind of like a forensic working copy. I’ve lost converted documents before, and have had to re-convert.I usually make a second file to edit from after that. Copy and paste the documents into the editing file. I’ll give GPT the increments from that file. I’ll place the increments into GPT with the following prompt: 1. Find: May I approach? may we approach? counsel, please approach. your honor, may we have a side bar? Counsel let’s have a side bar. Judge, can we come side bar? 2. place opening HTML head and body tags in the document. 3. Place level 2 headings whenever the words, “may I approach? or “may we approach?” or “counsel, please approach” or “may we have a side bar?” or “your honor, may we have a side bar? or “counsel, let’s have a side bar,” or “Judge, “can we come side bar?“in the document. 4. Finish the document by closing the head and body tags. I’ll take the GPT output and paste it into an HTML document named for that particular transcript. I’m thinking all this should work fine. However I’m not really sure. I mean, I tried to do this for a case in 2014 and failed because we didn’t have GPT, oh and it was only one 148 page document that the screen reader’s OCR program choked on. Although I’ve done them since then and it’s been fine. Here’s the prompt generator I used. hallanalysis.com/chatgpt-p… .