LINGUIST List 3.181

Tue 25 Feb 1992

Disc: Parsers

Editor for this issue: <>

Directory

1. Eric Schiller, Re: 3.175 Parsers
2. , RE: subject 3.175 Parsing and prosody
3. , parsing problems
4. Martin Haase, Re: Parsing challenges
5. , Re: 3.137 Queries: Like, Swahili/Yoruba, ESL

Message 1: Re: 3.175 Parsers

Since I am in the midst of constructing the rules for one of these
computational beasts, I find the discussion quite useful. Many of
these cases have deleted relative markers, so it seems sensible to
try inserting dummy relative markers after NP's when a parse fails.
Won't solve all problems, but perhaps it is a start.

With an Autolexically based approach, the semantics will have a better
chance of making it through, I think, though I haven't worked out the
details yet. The syntax is bound to gag without a kludge.

Eric Schiller

William Marslen-Wilson is surely right that prosody can provide some useful
information to listeners in parsing sentences. While I am dubious that it can
do much for "the player kicked the ball kicked him" , generally there is f0
(pitch contour trend and duration) information that can be useful. For
example in (1) I have put the approximate f0 at key points for my own
voice.

(1a) The professor (175Hz) the students believed was (175Hz)
	ar(150hz)rested died.

(1b) The professor (175Hz) the students believed was (150Hz)
	ar(175Hz)rested.

The crossing over of the pitchplots confirm what one's grandmother's ear
might have told us--the initial level of F0 is maintained over the potential
garden path deadend in (a). This could be used to preclude an immediate
(inaccurate) parse of "believed" as a main verb- either as a specific
"dependent clause" cue or--my guess--as a "procrastination" cue (Limber,
1976). I believe there's a paper by Grosjean in Cognition, circa 1988,
demonstrating that listeners can in fact use f0 to predict the subsequent
 length of an utterance.

Of course it is an empirical question as to the circumstances under which
these prosodic cues are applied in everyday speech; presumably William
Marslen-Wilson's experiments address this.
 John Limber

The recently-discussed examples of sentences containing "and and and and
and" and "had had had had had had had had had had" were both included in
Robert Ripley's _Big Book of Believe It or Not_, published in the 1930s.
I think there was an example with "that" also. Not to mention the six-page
sentence from _Les Miserables_ and the Comma that Saved a Man's Life and
palindromes and other such linguistic oddities.

 --David Bedell, U. of Alabama (dbedell3ua1vm.ua.edu)

Here is a parsing problem from French, at least if you want to deal
with its phonological transcription. French orthography has already
solved the problem:

Si six scies scient six cypre`s, 606 scies scient 606 cypr`es.
/si si si si si siprE si sa~ si si si si sa~ si siprE/

meaning: If six saws saw six cypresses, 606 saws saw 606 cypresses.

It shows that orthography is much more than a transcription; it is a very
efficient knowledge-based system.

Martin

Message 5: Re: 3.137 Queries: Like, Swahili/Yoruba, ESL

Sorry to be answering this so late, but I have been to busy
with our new baby to check out my e-mail lately! I don't
know any examples of exactly the "thrice repeated" type,
but I do remember a couple of similar examples that might
be of interest:

Dogs dogs dog dog dogs.
(Meaning: "Dogs who dog [= verb] (other) dogs dog [= verb] dogs."
Similarly:
Buffaloes buffaloes buffalo buffalo buffaloes.
(Meaning: "Buffaloes who (which?) buffalo [= verb] buffaloes
	buffalo [= verb] buffaloes."

Of course, neither of these makes eminently reasonable sse,
but then again, that's not the point.

As they say out here (and elsewhere???): "Enjoy!"

tom shannon
german department, uc berkeley
tshannongarnet.berkeley.edu