I see Roland mentioned a sniping tool for Windows. Modern versions of MacOS will recognize text from files (PDF and others) opened in Preview. If you open the PDF in Preview, you can select a text block, copy it, and paste it elsewhere.

There's also an indie Mac app called TextSniper that will allow you to quickly copy text from anywhere on your screen. It will even recognize text from a video -- something that still seems like magic to me. TextSniper is far more convenient than the Apple-supplied text scanner in Preview. The app is listed for eight bucks. It's a good tool from a good developer.

All of the methods will get some conversion errors.

Thank you, Roland, for your discussion. I tried your approach using a simple example from an online MMA book. As you can see from this small Notebook, it was not entirely successful. The problem is that TextRecognize does not understand the MMA symbol for Rule[] (and many other symbols which are unique to MMA:

https://www.wolframcloud.com/obj/e3fcc77f-d8a1-4fda-bb8e-1c91b18b2ec5

When I look at the raw source of the PDF page, I see that it switches fonts hundreds of times, sometimes on a character by character basis. All of the fonts are actually stored in the PDF (to guarantee portability), including some MMA-private fonts.

I'm in the process of writing code to parse the raw source and substitute an appropriate ASCII string when I encounter a symbol which uses an MMA font.

I will update this discussion with the results of this project.