The backstory does give some insight. A semi-automated solution will likely call for a developer capable of transforming the PDF text into a MARC-like text file that MarcEdit::MarcMaker (or similar) can then convert to MARC files that the ILS will load. That 2015 project mentions editing tools like notepad++ and sed.

Project Part 2 Using a Text Editor to repair the data
http://www.meanboyfriend.com/overdue_ideas/2015/07/worked-example-fixing-marc-data-2/