Used to lose ICONS information; why?

So I think in most cases where we used to not get ICONS in the result MRS in some of the regression tests, an argument-drop (e.g. head-opt-subj) is involved.

But I certainly did not rearrange anything in the type; I simply replaced diff-lists with append lists (you can click on the picture and the zoom in, I think):

Screen Shot 2020-07-27 at 10.21.23 AM

Why would a diff originate from that? Why were we not getting the ICONS using the type on the right and suddenly started getting it with the type on the left (the very bottom line of the MRS)?

To clarify, I think the diff is good, so, it is not a problem but I would just like to understand what happened. Why was the diff-list for ICONS underpecified with the type on the right?

Clicking isn’t providing a Zoom for me. Maybe the diff-list append was broken in some way that the append-list append isn’t?

Sorry about the zoom! Not sure why; it works for the second embedded picture…

Well yes, but would be good to understand what’s going on… That would have to be something fundamental, not having to do with this type, then. Unless something else changed on the lexical level (but I don’t think so? unless it was something subtle in the morphology library…)

Can you post legible versions of those types?

Actually, do you want to try and “open the image in new tab”? I think it’s big enough then.

That worked – and it looks like those types aren’t actually implementing the appends. What did the supertype for unary phrases that did the append look like for ICONS under the diff-list append?

I think you got it :). Used to get ICONS; now lost them; why? – the unary rule I think was doing the append in the reverse order :slight_smile: Maybe? see other thread; this seems to be the same issue anyway.

The order of the append shouldn’t matter unless something else is broken (leaky diff-list) somewhere…

1 Like

Dear Olga,

I agree with Emily’s thoughts. There seems to be a broken diff-list (or an append-list) somewhere.

Sanghoun

1 Like