ChatPDF – Chat with Any PDF

Post by darkteflon » Thu, Apr 20, 2023, 11:27 AM UTC

matchagaucho wrote: post #96

simonw wrote: post #20

How are you solving for PDFs that are too large to fit in the token context? I know of a few approaches for that: - Ignore the problem and let it hallucinate answers to anything that's not in the first 5-10 pages - Attempt to recursively summarize the PDF at the start - so summarize e.g. pages 1-3, then 4-6 etc, then if the resulting summaries are still too long for the context window run a summary of those summaries…

Chunk the PDF text and create embeddings. Get cosine similarity between user query and each chunk, and send the top N chunks to OpenAI that fit within token memory.

Post by burkaman » Thu, Apr 20, 2023, 1:17 PM UTC

zw123456 wrote: post #119

burkaman wrote: post #115

Earlier quoted context omitted.

That is a solid poem, is it actually a blackout poem based on whatever PDF you put in though?

Yes, I stuck a PDF of Godel, Escher and Bach, which is a book that my favorite prof in grad school gave me when I graduated back in 1980, of course the PDF is a much later version. But I thought it was pretty decent poem about the book. For anyone who doesn't know, a black out poem means it can only use words from the book or work you provide it.

Post by IKLOL » Thu, Apr 20, 2023, 3:34 PM UTC

mensetmanusman wrote: post #38

IKLOL wrote: post #3

How much is it actually using the PDF and how much is just normal Chat GPT knowledge? I uploaded a KJV bible and it seems to be doing pretty good with theological issues, like it knows salvation is by grace through faith alone which is my litmus test for any theological program. However, it seems to be just as honed in as Chat GPT is without even uploading a PDF.

If you look carefully at Paul’s writings, you will notice that he never says that our righteousness comes from faith alone—only that it comes from faith apart from works. Might need to wait for gpt5 for it to go more deeply into the topic :)

Post by stalfosknight » Thu, Apr 20, 2023, 5:19 PM UTC

taneq wrote: post #125

stalfosknight wrote: post #123

Earlier quoted context omitted.

It’s a piece of software guessing what text to spit out in a web browser. People need to get over themselves.

Is this something that people, in your experience, do?

Post by jzou1995 » Sat, Apr 22, 2023, 2:24 AM UTC

Docalysis wrote: post #89

simonw wrote: post #20

How are you solving for PDFs that are too large to fit in the token context? I know of a few approaches for that: - Ignore the problem and let it hallucinate answers to anything that's not in the first 5-10 pages - Attempt to recursively summarize the PDF at the start - so summarize e.g. pages 1-3, then 4-6 etc, then if the resulting summaries are still too long for the context window run a summary of those summaries…

I can answer for my site ( https://docalysis.com/ ) which does a semantic search to figure out which parts of the document are most relevant. Then you just use those parts. Docalysis also shows you the PDF side-by-side, has page numbers, and overall responses are of better quality according to users that have emailed comparisons to ChatPDF.