"Beware of the slippery slope of messages"


#1

I went through every mention of “messages” that I found on the DELPH-IN wiki, and all of them postdate the actual discussion of why messages were eliminated (apart from this but it is not very informative). Is there anything else I could look at, or could someone summarize it here?


#2

I can’t answer your question directly (messages were before my time, even), but I think in general the DELPH-IN wiki is in pretty bad shape. Lots of information is outdated and/or incomplete, and it’s also generally hostile to outsiders (i.e., useful for people who have already been sufficiently indoctrinated, less so for neophytes or heretics (“muggles”, if you prefer)). If only we could work “wiki maintenance” into a budget somehow…


#3

I think notes from discussions, however outdated, can be very valuable, it is just that in this case, there aren’t any relevant discussion notes…

(But I don’t disagree that it is not always easy to use the wiki, of course, you know I of all people would not do that :wink: And the neophyte problem is exactly on point though I think solving that problem is more akin to writing a textbook (“lecture notes”) than to maintaining a wiki…).


#4

I think notes from discussions, however outdated, can be very valuable

True, but I think the historical content should be clearly labeled as such.

it is just that in this case, there aren’t any relevant discussion notes…

That’s unfortunate. You might also try grepping the LKB source tree (I’m not joking… I’ve learned quite a bit from the comments (if not the code) in the Lisp sources), or see if the term is mentioned in any papers (e.g., Google Scholar) or in the mailing list archives (not just developers… here’s one from the erg list: http://lists.delph-in.net/archives/erg/2007-March/000007.html, and the matrix list: http://lists.delph-in.net/archives/matrix/2007-December/000089.html)


#5

On wiki maintenance and the value of notes: It seems to me a little mean-spirited to complain about the wiki in this way, given that in most research collaborations the conversations documented in those notes would have been completely ephemeral and not documented except in as much as they eventually informed publications. Please adjust your expectations of the wiki accordingly.


#6

Right, well, the point of this is not complaining about the wiki, it is more about how I find out more about what the “plethora of problems” (a quote from the wiki!) was with messages :). The reason I even mention the wiki in the first place was to prevent people from suggesting I go and read the wiki :).


#7

On the history of messages, here’s what I can recall:

We (= me, originally) put them into the ERG in an attempt to model the notion of ‘message’ in Ginzburg & Sag in MRS terms. The point was to capture the differences of sentential force that are syntactically marked, so questions v. commands v. statements, and possibly finer-grained versions of those. In G&S this is the ‘container’ structure inside of which you find the rest of the semantics. The message rels were a flattened, MRS-style way of doing that.

The problem is that once we started putting them in, it wasn’t clear exactly which linguistic elements merited their own ‘message’. Finite main clauses, for sure. Finite complement clauses pretty clearly too. Non-finite embedded clauses? Relative clauses? Reduced relative clauses? Non-clausal (but still event-bearing) modifiers of nouns?

On top of that, we found (or maybe imagined, or maybe both) that they were off-putting to outsiders we were trying to convince to be interested in MRS — especially if we were generous in inserting them, they added to clutter in the representations.

So, Dan took them out of the ERG, and instead posited a feature on event variables (called SF in the Matrix) and the other grammars and the Matrix followed suit. SF, being a feature on events, is there in all EPs with an ARG0 of type event, but it can be left underspecified without fuss.

Does that help?


#8

Yeah, that remark was more for Mike, I think.


#9

Excellent, thanks, Emily!


#10

@olzama I apologize for hijacking your post to rant about the wiki

@ebender I don’t mean to disparage the work of wiki contributors or underappreciate the value of what’s there, and “hostile” was too strong a word. Also, rereading my post it seems “that’s unfortunate” could sound like disapproval or accusation, which was not my intent. I meant that the wiki could be much more useful with some maintenance but unfortunately we rarely have time to devote to the task. But this was not the right place for me to start such a conversation.


#11

Thanks, Mike!