home
events
contacts
mailing list
directions
Linguistics
Department
Stanford
University |
Stanford
Humanities Center
Mellon
Foundation
Graduate
Research Workshop Program
Stanford
Semantics and Pragmatics Workshop:
THE
CONSTRUCTION OF MEANING
SemFest, March 14, CSLI:
16:00-16:30
David Oshima and Roger Levy
Nouns with multiple classifiers and non-transitive information flow
in Japanese
In Japanese, as in other classifier languages such as Chinese and
Malay, numerals usually do not directly quantize common nouns, but
must first combine with a classifier (Aikhenvald 2000, Gunji and
Hashida 1998). The selection of classifier is based on the semantic
properties of the noun to be quantized. From the perspective of
constraint-based approaches to syntax/semantics, the selectional
restriction between classifiers and nouns can be stated in terms of
the unification of sematic features, to some extent parallel to the
treatment of gender/number agreement (between determiner and noun, for
instance) (cf. Kathol 1999).
In at least two cases a single noun can be associated with more than
one type of classifier. First, virtually any noun can be associated
with the "taxonomic" classifier `shurui' in addition to its sortal
(individual- or quantity- measuring) classifier(s). Second, certain
polysemous nouns can be associated with more than one type of sortal
classifier, according to what aspect of their semantics is
"highlighted".
It can be shown, however, that this one-to-many correspondence between
nouns and classifiers is not a problem of ambiguity. In Example 1
below, the classifiers `satu' (volume) and `bu' (copy) are both
associated with a single occurrence of the noun `hon' (book). This
could not be captured by an ambiguity analysis.
1. Sono shuppansha-wa eiga-ka-s-are-ta 2-satu-no
that publisher-Top make_into_a_movie-Pass-Past 2-CL.volume-Gen
hon-o 2000-bu-zutu zoosatu-si-ta.
book-Acc 2000-CL.copy-respectively reprint-Past
`That publisher printed 2000 additional copies of two books which
were made into a movie.'
The ability of a single noun to support multiple classifiers also
differs, however, from cases of syncretism as discussed by Zaenen and
Karttunen (1984) and many others. The reason for this can be seen in
Example 2 below. The noun `basu' (bus) is compatible with two
different sortal classifiers: `dai' highlights it as a physical
vehicle, whereas `hon' highlights it as a certain type of scheduled
event. The distinct semantic highlightings available for `basu' are
compatible as participants in distinct types of events. In 2a, the
verb-object combination `ziko-ni makikomu' (get in an accident) is
compatible with the conceptualization of `basu' as a physical vehicle
but not a scheduled event. In 2b, the verb `nogasu' is used in the
sense of missing a scheduled event and is not compatible with `basu'
as a physical vehicle. In 2c, however, the presence of the verb
`nogasu' in a relative clause does not prevent the measuring of `basu'
with `dai' in the matrix clause. This suggests that restrictions on a
noun's sortal type can be resolved distinctly in different clauses.
2a. Basu-ga ziko-ni 1-dai/*pon makikom-are-ta.
Bus-Nom accident-Dat 1-CL.vehicle involve-Pass-Past
`One bus got involved in an accident.
2b. Basu-o 1-*dai/pon nogasi-ta.
Bus-Acc 1-CL.event miss-Past
`(I) missed one bus.'
2c. Boku-ga nogasi-ta bus-ga ziko-ni 1-dai/*pon makikom-are-ta.
I-Nom miss-Past bus-Nom accident-Dat 1-CL involve-Pass-Past
`One bus that I missed was involved in the accident.'
To give a proper account of the multiple-classifier example in 1, we
propose to enrich type-feature logic with dotted types and an
operation to construct them (cf. Pustejovsky 1995). This allows us to
solve all the unification requirements in 1a (i.e. mutual restrictions
between the classifier and nouns, and semantic restriction posed by
the predicate on its arguments). To account for the partial sortal
restriction conflicts seen in 2, we argue that information-sharing
between verb, noun, and classifer is not completely transitive, and
present two alternative formal analyses capturing this: one based on
weak unification (cf. unificability tests as proposed by Ingria
1990), and the other based on modeling multiple selectional
requirements with consecutive set intersection.
REFERENCES
Aikhenvald, Alexandra Y. 2000. Classifiers: a Typology of Noun
Categorization Devices. Oxford University Press.
Gunji, Takao and Koiti Hasida (1998) ``Measurement and
Quantification'' in Takao Gunji and Koiti Hasida (eds.) Topics in
Constrained Grammar of Japanese, Kluwer: pp.39-79
Ingria, Robert J. P. 1990. The Limits of Unification. Proceedings of
the 28th Annual Meeting of the ACL.
Kathol, Andreas. 1997. Agreement and the Syntax-Morphology Interface
in HPSG. Ms.
Pustejovsky, James. 1995. The Generative Lexicon. MIT Press.
Zaenen, Annie, and Lauri Karttunen. 1984. Morphological
Non-distinctiveness and Coordination. Proceedings of ESCOL 1984.
Please contact one of the workshop organizers
if you have suggestions for presentations or the workshop in general.
Back to the workshop homepage.
This workshop is sponsored by
the Stanford Humanities Center, and funded by a grant from the Mellon
Foundation.
|