Going to the USPTO website and typing
aclm/(Schematron) into the advanced search field shows US patents that mention Schematron in their
actual claims (the core of the patent). Here is the result:
I have briefly gone through the patents, and here are my preliminary comments. US Patents are frequently written so that the novel stage is buried down the list, or generalized to reduce the chance of having gaps that some other tricky patent can take over. So it is quite possible that I have missed the innovative step. I expect I will look over some of these in more detail. While I think some are silly (#7) and some do not have any actual techniques (#6) it is only #1 that gets my goat as an obvious idea that encrouches on Schematron.
The system may do some NLP to help. The system converts the unified document into Schematron, XSLT, Java etc. The system runs the tests and combines the boolean results, and formats and collects the formatted fragments.
There are 50 other patents that mention Schematron, but in the specification rather than the claims.
I see IBM continues to make untrue statements about Schematron: this time in patent
9,329,860 they claim that Schematron cannot express that all elements are supposed
to be optional. They conflate Josh Lubell’s Schematron Design Rules with Schematron
Rules themselves, I think. But Schematron can easily say for example
<assert test="count(*) = count(A) + count(B) + count(C)">This element can contain
only A, B and C elements, but all are optional</assert> .
Patent 8,214,421 is utterly nutty, as far as I can see. It is a patent on using validation to check conformance to a standard rather than comparison against a reference implementation. It is now 40 since SGML was standardized, enabling testing of documents against a schema (DTD). Yee Gods!