Tell HN: Here is a hard problem for you all that want an idea

It is 2018, we have self driving cars, but for some reason, we still cannot get printed material digitization to work flawlessly.

As it is done currently, the book is scanned, then OCRed, the digital book is then assembled, from the scanned images with a hidden layer of OCRed text, so that you can select and search for text in the digitized book.

Why this sucks:

1. readability is pretty poor, because you are reading a scanned image on a screen.

2. OCR is pretty poor. It can only recognize basic alphabet. Any mathy book will have math all over that is recognized as giberrish. This is the reason why they cannot construct the digital book from the OCR data alone

3. in the far future one may dream of such incredible software that would convert a scanned mathy paper into a well formatted latex document; but that is too much to ask for now

----------------------------------------------------------------------

I know this is actually a hard problem. But it is definitely a problem we can solve with today's tech. It is just a big project.

I do know that some entities do offer proprietary solutions. I have not personally tried them because I am not the one digitizing books. I don't know how good those solutions are, but most digitized books I come across are pretty poor quality, still.