What good is Syntactic Information in the Lexicon of a Syntactic Parser?

Imagine a situation where we want to parse texts and get as output information about the traditional grammatical categories of senten­ ces, i. e. subject, direct and indirect object, finite verb, ad­ verbials etc. That is an explicit goal of most parsers, and a necessary prerequisite for many applications of computational linguistics. How much information, syntactic and/or other, do we need in order to reach that goal?