Yes, using the LaTeX source code (or HTML, once that becomes reliable and widely used) should be much more robust than PDF parsing.