booze-tools VS tika-python

Mmmm... Indirectly... Sophie uses my literate parsing system.

It's not my idea. It's present at least as far back as YACC, "Yet Another Compiler-Compiler", which inspired the name of BISON (another parser-generator named for ungulates). Here's mine, written in Python: https://github.com/kjosib/booze-tools It also has a few extra bits. Feel free to exploit its MIT license to the fullest. I should mention that the design of symbolic reduce-actions was intended to allow one to use the same grammar across multiple host languages. You could even write a driver that does simply build a parse-tree and then hand that off to a separate phase, but in my world I almost always want a bottom-up tree-transduction as first-pass de-sugaring.

Many CS schools have an undergrad class called "compilers" in which you'll implement (from the ground up) either a Scheme or a thing-that-is-like-Java called decaf, or possibly you'll implement Scheme and just call it decaf. If your school does not have such a course, you can get class notes from Texas A&M or Stanford or a variety of other places. (Here's a parser for it.)

At my previous job we had the same problem which we solved by using Tika. We called it on the server along with other stuff, but there is also a Python binding.

Tika is from Apache so yes its original code base is Java but it has bindings in other languages. Checkout Tika-Python!

The Textractor instance is the main entrypoint for extracting text. This method is backed by Apache Tika, a robust text extraction library written in Java. Apache Tika has support for a large number of file formats: PDF, Word, Excel, HTML and others. The Python Tika package automatically installs Tika and starts a local REST API instance used to read extracted data.