I'm sure it's not what you want to hear but basically you can't do this. At least not without insider help at Microsoft. In order to facilitate this, you would need to develop your own language recognizer. And as far as I know, the tools and documentation for doing so are not provided to third parties.
However, you could develop your own recognition system that utilizes the existing ink collection and stroke model provided by the System.Windows.Ink
API's, but it would not plug into the InkAnalyzer
model. Utilizing the stroke information will yield much better results than a simple OCR type of system where you're simply analyzing shapes.
It would have been very useful for Microsoft to expose their math parsing capabilities in the System.Windows.Ink namespace but as that was coming out, Microsoft was giving up on tablets. Then the iPad came along and ate their lunch and now who knows if we'll ever see a focus on ink input again.