I have to admit, PDF parsing being such a hot and profitable topic in computer science was really something I never saw coming.
PDFs? The things you can select text from? And when not, there’s decent OCR? And when not, you just ask the person to send you an email or a word doc?
It sounds like LLMs are looking for a new unpolluted source of historical data that they can learn from, and this source exists in the form of old scanned-in paper documents. That’s the only reason I can fathom as to why this is such a big thing now.
Every time I try to convert a PDF to epub or something, or OCR one that doesn’t actually have selectable text, it turns out shit. I assume the real reason people would want to get LLMs involved is that there is actually a lot of ambiguity in what a correct conversion would be, and there are a lot of PDFs out there.
I have to admit, PDF parsing being such a hot and profitable topic in computer science was really something I never saw coming.
PDFs? The things you can select text from? And when not, there’s decent OCR? And when not, you just ask the person to send you an email or a word doc?
It sounds like LLMs are looking for a new unpolluted source of historical data that they can learn from, and this source exists in the form of old scanned-in paper documents. That’s the only reason I can fathom as to why this is such a big thing now.
Every time I try to convert a PDF to epub or something, or OCR one that doesn’t actually have selectable text, it turns out shit. I assume the real reason people would want to get LLMs involved is that there is actually a lot of ambiguity in what a correct conversion would be, and there are a lot of PDFs out there.
Training the most insane AI model on classified federal documents.