Olga Zamaraeva

I started working on this, and for now I just have this (but, assuming it is working correctly, this is already helpful): ``` NOTE: parsed 7 / 16 sentences, avg...

The above example is saying that the only check that failed was the "simple tests" check for the length of ICONS and such. If any other checks also failed, there...

I tested this a bit and am now sure that what I have does not work correctly. For example, I changed *transitive-lex-item* to not actually link the ARG1 and ARG2,...

Ah, that was a very simple bug. So now I get somewhat uninformative messages which are nonetheless indicative that something failed while comparing two predications, rather than the length of...

And looks like I can easily make it somewhat more(?) informative like this: ``` ******** Messages from MRS comparison ********* 1: Length comparison failed between e2 and e2. None comparison...

Nah, I think that's as uninformative as before... e2 seems just to be some kind of top structure.

Or maybe it does work, kind of (btw I hope it is OK that I am talking to myself here?) Below, I broke specifically something in a noun relation, keeping...

(3) is the only thing I really need, with the caveat that if two different mechanisms are used to compare MRS to report an rtest failure and to find out...

To talk a bit more about (1) and (2) though, do I understand right that what's intractable is in fact distinguishing informative diffs from uninformative? Using vf2.

Gotcha; that's what I thought. In that case, I think ideally, we would have a faster rtest and a slower rtest. If the faster one returns a failure for a...