How to parse weakly formalized text?

Question

Machine learning

There is a need to parse this list of drugs with descriptions and a list of drug synonyms.
As a result, I would like to see a json file with a code like this :

Using dom, css selectors and regular expressions, this cannot be done (more precisely, it does, but only partially), because you have to write rules for a very large number of exceptions in the form of typos, nested brackets, nested lists with different string formatting, etc.