So the Good Tables web service seems to have a way to test a schema against a CSV file, with the possibility of changing the schema to more accurately describe the CSV File if it isn't a good match.

I'm not really seeing anything on the related YouTube videos that shows this uses any technology stronger than a loop and some regular expression parsing. I thought Open Knowedge was advocating the use of ontologies and artificial intelligence processes such as rules engines.