Hi There,
I’m using the Mac M1 erg-2020 version of ACE from this forum Compiling ACE on MacOS - #26 by EricZinda, and had a question about how the ERG is intended to represent phrases with “else.” For example, when I parse the phrase, “what else is on the menu,” I get two parses:
[ “what else is on the menu”
TOP: h0
INDEX: e2 [ e SF: ques TENSE: pres MOOD: indicative PROG: - PERF: - ]
RELS: < [ thing<0:4> LBL: h4 ARG0: x3 [ x PERS: 3 NUM: sg ] ]
[ which_q<0:4> LBL: h5 ARG0: x3 RSTR: h6 BODY: h7 ]
[ _on_p_loc<13:15> LBL: h1 ARG0: e2 ARG1: x3 ARG2: x8 [ x PERS: 3 NUM: sg IND: + ] ]
[ _the_q<16:19> LBL: h9 ARG0: x8 RSTR: h10 BODY: h11 ]
[ _menu_n_1<20:24> LBL: h12 ARG0: x8 ] >
HCONS: < h0 qeq h1 h6 qeq h4 h10 qeq h12 > ]
The first parse doesn’t seem to have any representation of “else” – in fact the same parse can be achieved from “what is on the menu.” The second parse, however, include “ellipses_ref.” How is that ellipses ref intended to represent the semantics of “else?” Thanks
Your observation points to a shortcoming in the ERG’s analysis of “what else”, where the word or phrase following “what” fails to contribute any additional semantics. You’ll see similar behavior for “what more”, “what exactly”, and even “what the heck”. I guess “what else” should have a meaning something like “what additional thing”, so I’ll make that improvement, but it’s less clear what the semantics contributed by “the heck” should be. Similarly for “what on earth”. Suggestions would be welcome.
As for the second parse with the ellipsis_rel, it is unrelated to “what else”. In this analysis, the “be” is analyzed without its complement, then modified by the PP “on the menu”, pragmatically unlikely but syntactically well-formed. Consider the following embedding: “I see one dish that may be vegetarian at this restaurant, so on the menu, what else is?” It’s true that conventionally, this second reading would be punctuated with a comma as in “what else is, on the menu?” corresponding to the contrast in intonation for the spoken utterance, but that comma is not always reliably provided, so the grammar allows for the elliptical reading even without, expecting the parse selection model to rank it as less likely when there is a better alternative reading.
Thanks so much-- it’s interesting to think about how in construction like “what exactly is that?”, the “exactly” contributes semantic information to the entire sentence, yet it’s difficult to pin down how and where that information is applied.