MRS with messages

(Related to this)

Here’s an MRS from ERG version 1143 (“last message-ful version”), for the sentece Who left.

Screen Shot 2020-07-28 at 11.45.40 AM

The basic_int_m_rel and the prpstn_m_rel are he message relations, right?

So the MRS is saying: there is a question about x4 and this question is a parameter of e2, something like that?

It looks like we don’t have anything analogous to the Ginzburg & Sag PARAMS list there — but the wh parameter is showing up as the TPC (topic) of the question message (basic_int_m_rel). On the other hand, it’s also the TPC of the proposition message (prpstn_m_rel) to TPC can’t have been a stand-in for PARAMS.

The ARG0 of the messages is shared with another event in the sentence. For Who does Kim think left:

Screen Shot 2020-07-28 at 12.02.07 PM

The interrogative message relation shares its ARG0 with the matrix clause event variable rather than with the embedded one. The embedded one has a propositional message associated with it.

I’m not exactly sure what the ARG0 of the message was supposed to represent, but given that it’s there and identified with something, it seems to make sense that it matches the clause the message is associated with.

So, for the purposes of writing about this, is the fact that the ARG0 of the interrogative relation is associated with the matrix clause, and the TPC is the ARG1 of the embedded clause, sufficiently illustrative of what we used to do and no longer do, then? I don’t know how important it is to link this directly to the PARAMS feature in G&S…

Summary of conversation with Emily: in the above, the ARG0-s seem to sort of be doing duplicate work wrt HCONS; at any rate, they are certainly not the same thing as the PARAMS feature in Ginzburg and Sag. That would’ve been an additional feature in these same message relations, probably. Seems like it never was added to ERG releases.

The basic_int_m_rel is associated with the matrix clause rather than the embedded clause (its ARG0 is e2 rather than e19), and it points to the wh-element (its TPC is x3). It looks like TPC is a single individual rather than a list, but otherwise it’s tempting to see it as PARAMS. The link between e2 and x3 is exactly what we’ve been saying we need.

It’s not the same as a PARAMS list (as @ebender pointed out, the TPC is also on the prpstn_m_rel), but based on this example we could easily convert the TPC to a PARAMS list of length one. But maybe this doesn’t work elsewhere – I’d be interested to see what the TPC is in other cases like “Does Kim know who left?” or the examples in Question representation (examples from the ERG?).

I’m fairly sure that TPC (and PSV) were introduced to try to model information structure and not wh parameters. I have a vague memory of trying to build a PARAMS list in the original message implementation (for which I take the blame) but I think it was either never completed and/or whatever I did build wasn’t included in the ERG trunk.

I see. So then a parallel between TPC and PARAMS would be incidental.