Thanks for your question and comment James. In my case, I took 235 Brazilian FOMC minutes, and you can tell several terms and words are standard (if there was ChatGPT in the 90s, I could swear it built those minutes, lol). Still, AI could filter well those words as "meaningless" and find significance in no more than a few words.They showed surprisingly high sentiments and, worse, seasonality.Keep up with your research and pursuing alpha!