It seems that there may be an issue with Tern pulling text out of these documents. You can try using original documents before they were converted to PDFs to see if that works better. Sometimes the way PDFs are created can cause problems for the AI parser.