noun-entity related conceptual problem in test sentence

[JosephNathaniel]

One of the manual test sentences Alice likes what she knows raises an interesting conceptual problem, again related to the question of wires/referents.

After pulling out (but before type expansion), the diagram looks like this:

Applying n_type_expand to this does nothing, which is what the code is designed to do in this scenario.
However, this causes something strange, because the she wire does not come out at the end and cannot be referred to.

Indeed, after s-type expansion, we get

Note: this will also lead to weirdness in the intra-sentence coreference, since we want to identify Alice and she at the top, but at the bottom the two wires correspond to Alice and what Alice knows, which should not be identified

[RazinShaikh]

In a correct diagram, "likes" should be inside the "what" frame and it should take in the "knows" wire. However, at the end, the type of the wire that comes out should be "she", not "knows". But I don't know how to explain that starting from the ccg parse of this sentence.