Comment by kol
10 years ago
Great idea! It'd also be great if someone created hackernewspapers.com too. There are so many great papers mentioned on HN. Since most links point to pdf files, it'd be possible to collect the pdfs on the same site (if there's no copyright violation).
That can be done, question is how to recognize id a pdf is a paper or not. A bit trickier than books imho.
That sounds like a fun application of a neural network. Finding large repositories of "real" papers to train it with should be fairly easy, too.
That said, I have zero experience working with neural networks, so it might not be very easy at all.
You'd get pretty far just by whitelisting arxiv, .edu, .ac.uk, etc.