Handle multiple sentences more elegantly

[lynn]

Right now, given input like Dıo da. Nuaq da., Kuna returns a faux syntax tree like [Discourse [SAP ...] [SAP ...]], but I would prefer for the output to be more like an array of individual trees, that are (for example) browsable with "next" and "previous" buttons in the web interface.

---

Kuna's purpose is to analyze syntax, and the largest unit of syntax is a sentence. So Kuna's purpose is not to analyze discourses. Even if there's a sá shamu in one sentence and shámu in the next, it's fine for Kuna to handle the shámu in the second sentence as bound exophorically.

You can imagine that Kuna doesn't "know" if the second sentence was even part of the same conversation, or how much time was between the sentences, or generally what the relation between sentences is, who's talking, etc. It just sees one sentence at a time. This makes our job easier, but I think it's also justified linguistically — discourses don't have the same "binary branching" structure that sentences have (in the Minimalist program), so it doesn't make sense to try to extend the Merge structure upward.

In the words of the refgram (on :d2:):

If a vehicle was mentioned, then 󱛘󱚿󱛊󱚺󱛎󱛃󱛙 cháo probably refers to that vehicle.

The key word is "probably" — we can be conservative in Kuna and assume the weakest (i.e. no) relationship between input sentences.