I found an intresting solution for classifying keywors are already classifyed by a language service. It's description says it uses a Tagger to enhance code highlighting. Maybe it can help you: KeywordClassifier Older version of the linked project used a classifier mentioned in the description.
You can get the name of the loaded document, also the extension with ITextDocumentFactoryService or maybe there is a way to bind the tagger also to extensions not only to the content type of Basic (instead of code
). FileExtensionAttribute may help.