Those are good questions. The problem isn't really how you represent them in a tree, but what sort of analysis you'll choose.

A N-N compound might be represented as (at least):
1. An Adj-N structure, with the first N coerced into being an Adj.
2. An NP with an internal NP modifying it (in place of an AdjP).
3. A N-N structure (as you have above), although that's a little strange (each phrase should have one head, and each head should have one phrase-- this doesn't match).
4. Part of the morphology-- it's a compound word, not syntactic.

There are arguments for all of those. Which one you pick determines which one you use in the tree.


The second case represents the "double object" construction in English (search for that to find more info). The right analysis is not clear because it doesn't appear to be binary branching. There are a few ways around that. One is to assume this is an exception and have ternary (three-way) branching for this structure (possibly like with conjunctions?). Or you can attempt to modify the structure slightly so it fits-- maybe there's movement (from an original structure that is more normal) or maybe there's a hidden element. One option for a hidden element would be a null preposition ("baked Mary [Ø pie]" where Ø is vaguely something like "with").

In other words, for both constructions, there are a few options. The starting point would be constituency tests to see how these words behave as groups. But each of these represent generally unsolved problems that might be analyzed different ways by different researchers. That doesn't mean you can't adopt one analysis or another and use it effectively; it just means there isn't an easy/standard answer to your questions.