Gosse Bouma (gosse@let.rug.nl)
Mon, 20 Jan 1997 22:26:29 +0100 (MET)
All, Ivan writes: > As this exchange about adverb extraction has progressed, we seem to be > copying more and more people, so we decided to move it to the hpsg > list. Bob Levine and I were also wondering why we don't have more > discussion of this kind on that list. It's not that people aren't > working on issues like this that are of common interest... I got to read this a day before the rest of you. Does this mean I am part of the incrowd, or does it mean that by the time Ivan decides that he has to cc to some Dutch guy, it must be a matter of common interest? [long story leading to a proposal for an Adjunct Introduction LR omitted] > > ------------------------------------------------------------------------- > Bob then replied: > Date: Sat, 18 Jan 97 14:16:43 PST > From: "Ivan A. Sag" <sag@csli.Stanford.EDU> > To: sag@skol.Stanford.EDU > ------------------------------------------------------------------------- > > >Ivan's new rule does avoid the problem of enforced narrow scoping > >that caught my eye in the last version of the rule: once you've added > >the adjunct synsem value to the COMPS list, you can then run the > >output of the rule through the CELR, and since the output of the > >adjunct LR will be a head whose CONT reflects widest scope for that > >added adjunct, the output of the CELR will allow you to extract that > >adjunct while yielding a head whose CONT also reflects widest scope > >for that (now extracted) adjunct. And it also avoids all the other > >technical problems. But there's something else about it... isn't this > >essentially the same rule that's presented in the van Noord & Bouma > >paper on adjuncts as complements that came out a few years ago in one > >of the computational journals? I'm *sure* it's basically that same > >rule! A rule like this has been proposed by several authors. Philip Miller proposes something like this in his dissertation, Iida, Manning, and Sag made this proposal in a paper on Japanese causatives (later revised into a `zero derivation stem type') and there is our proposal. Actually, we were inspired by some unpublished CG work by Jack Hoeksema, who also refers back to someone else, if I remember correctly. I also seem to remember a CG paper where it is argued that coordination provides evidence for letting heads select their adjuncts (not by means of type raising, but by means of type polymorphism, Barry and Pickering ? Hepple ?) Interestingly, both our and Iida et al.'s motivation for putting adjuncts on comps had to do with scope in the context of complement inheritance. In examples such as (1) below, there is an ambiguity: ...dat Arie Bob de vrouwen met een telescoop zag bekijken ...that A. B. the women with a telescope saw look-at that A saw B look at the women with a telescope There is good reason to assume `zag' is a complement inheritance verb. However, if `bekijken' never combines directly with its complements, the narrow scope reading seems impossible to derive UNLESS adjuncts are on COMPS. (There is a brute force alternative of course: if adjunct scope is determined by the same mechanisms which account for Q scope, configurational considerations loose their force.) [long story about French adjuncts omitted] > > > >So it turns out (at the moment anyway) that AGKMS seem to be requiring > >at least unicity and some evidence of variation in lexical properties > >on the head in order to qualify an adverb as a complement. Evidence of > >sisterhood to the head is insufficient, since that can be accomodated > >without positing complementhood. More to the point, AG evidently do > >not regard extractability of adverbs as evidence of complementhood; > >French adverbs extract just as English ones do, but that apparently > >wasn't taken by AG as evidence in favor of complementhood. Was this an > >oversight, or was it based on the sense that the most important > >consideration was the lack of evidence for a `split' in a verb's > >lexical entry when a non-negative adverb is present as vs. when it > >it's not? If it's the latter, then the approach that Gosse and Gertjen > >first proposed in their paper and that Ivan and Rob rediscovered seems > >to run into the same problem as the French non-negative adverbs: the > >LR rule in question coins a new avatar of the verb, but where is the > >evidence that you really have a separate lexical item? I would like to reverse the argumentation here: unless there are arguments which show that adjuncts must not be selected by the head, let us assume that they are: 1. This accounts for scope in the context of complement inheritance, and adjunct extraction (as observed in Ivan's contribution). 2. It seems to me that this is practically the only way to account for adjunct extraction in the context of Ivan's recent approach to (complement) extraction: In Ivan's Constraint-Based Extraction paper, the NLFP is replaced by a lexical constraint, slash amalgamation, which defines the slash value of a head as the set-union of the slash values of its ARG-ST members. The Adjunct Extraction LR as formulated by Ivan creates a rather brutal exception to this rule by inserting an element on SLASH which does not correspond to an ARG-ST element. Note furthermore that if we need a LR to put adjuncts on COMPS (and ARG-ST) anyway, the Adjunct Extraction LR becomes superfluous (since adjunct extraction is now a special case of complement extraction). The introduction of ARG-ST makes the issue even more complex, of course, as it could be imagined that adjuncts are on COMPS, but not on ARG-ST. This might account for the binding data observed by Levine: > >(1)a. *You can't tell them{i} that the twins{i} are being offensive. > > b. You can't say anything to them{i} without the twins{i} getting offended. etc. However, in Iida et al., adjuncts are in fact added to ARG-ST. In another one of these discussions which ended by it being cc'ed to lots of people, Chris Manning observed: > It depends on one's theory of binding > in adjuncts (which, admittedly, has never been very well developed), but if > one takes the simplest theory that a (long distance, Principle Z, > non-exempt) anaphor in an adjunct must satisfy the same condition of being > a-commanded as other anaphors, then adjuncts have to be on arg-st. For > example one would need it for a sentence like the following (which I just > invented, so I think it's grammatical, but it may well be awkward): > > Mitio ga zibun o hihan sita ato de taihen kanpai sita > Michio NOM self ACC criticize after great.deal worried > 'After Michio_i criticized self_#i/j, [I/she/...]_j worried a lot.' > > A second argument to the same effect could be made about the controlled > subject of -nagara clauses -- mentioned briefly in MS&I p.26. Dutch `zich' provides a similar case: (i) * Zij_i vergaf zich_i nooit she forgave ZICH never (ii-a) Eva_i zag mij plotseling voor zich_i staan Eva saw me suddenly before ZICH standing (ii-b) Anton_i verwacht met Simone naast zich_i het proefwerk Anton expects with Simone next-to ZICH the test te kunnen maken to be-able-to make (iii) *Jij zag mij plotseling voor zich_i staan you saw me suddenly before ZICH standing At least superficially, `zich' and `zibun' have similar properties (i.e. the must be free in some minimal domain (as opposed to anaphors), but must be bound in a larger domain (as opposed to pronouns). The problem here is not only that there is no well worked-out theory of binding in adjuncts, but also that the idea that binding applies to ARG-ST has never been considered in detail. It means at the very least, I guess, that all referential and anaphoric elements must be on some ARG-ST (what if they are on two, as seems to be possible in some proposals?), and that these should all be in some kind of `command' relation to each other. Any suggestions? Gosse. -- Gosse Bouma, Alfa-informatica, RUG, Postbus 716, 9700 AS Groningen gosse@let.rug.nl tel. +31-50-3635937 fax +31-50-3636855
This archive was generated by hypermail 2.0b3 on Fri Dec 18 1998 - 20:36:03 PST