So some days ago I asked in this community what kind of AI model should I use (and how could I use one) to classify several email replies that I had from scientists after asking them a few questions to them. I finally paid for Perplexity pro service and it apparenly did a nice job classifying them. I finally gave the model the PDF with the actual answers from the addressees and another PDF with the "expected answers", and asked it to count the number of answers that overall coincide...