Bernard Lamers' Research Homepage
Last updated: March 25, 2003
University of Tokyo
7-3-1 Hongo, Bunkyo-ku, Tokyo
Ph. D. Student
(M.A. in Japanese,
M.Sc. in Interdisciplinary Information Studies, specialty Computational
Please replace 'nospam' with 'iii.u-tokyo.ac.jp' before sending mail.
- The syntax-semantics interface, especially its realization in
- The limits and possibilities of a mixed linguistics/statistical
approach to language modeling.
- Deployment of linguistic theory in actual NLP systems
Publications under preparation
The inadequacy of the Double-o Constraint: abstract
Japanese marks nouns with particles to indicate their grammatical
function. The relationship between particles and grammatical functions
however is not exclusive: a particle may be used as marker for more than
one grammatical function, and a grammatical function may be expressed by
more than one marker. This results in sentences in which one marker
occurs more than once. The particle ga for example, usually a
subject indicator, may -depending on the predicate- also be used as
object indicator, as in Boku ga susi ga suki da. (I like
sushi). The particle o can be used as object-marker or
causee-marker. We would thus expect constructions like: *Boku ga
Takasi o musi o tabesaseta. (I made Takashi eat an
insect). Surprisingly, this sentence turns out to be ungrammatical. In
order to account for this fact, the Double-o Constraint was
defined. This constraint forbids multiple occurences of o-phrases
within a single clause, and hence correctly rules out the earlier
The Double-o Constraint has become a de facto standard in the
analysis of the Japanese causative construction. Unfortunately, there
does not seem to be any motivation for this constraint in constructions
other than the causative. Even worse, the Japanese causative
construction is so complex, that the Double-o Constraint alone
will not be helpful in explaining the data. This paper will show that at
least two more constraints are necessary to successfully capture the
Japanese causative data.
An analysis of the causative construction based on three constraints
however is neither very elegant nor insightful. This paper unfolds a
completely new analysis of the Japanese causative, partly basing itself
on particular behavior of the causative which no paper addressed so
far. The theory presented provides a clear, yet powerful mechanism to
explain the whole spectrum of causative data. Also, being firmly rooted
in a mapping mechanism of semantic features to syntactic particles, this
approach provides further support for the idea that semantics guide
syntax, and syntax reflects semantics.
This paper was presented at the March 22, 2002 meeting of the
Semantics Group of Tokyo University and also at the December 5, 2002
meeting of the Linguistics Group of Waseda University.
Reference material for presentations held at the
Nakagawa/Sugimoto study sessions
- Research presentation held at April 24 and May 1,
Paper discussed: Merlo, P. and Stevenson, S. 2001. Automatic Verb
Classification Based on Statistical Distributions of Argument
Structure. Computational Linguistics 27:3
pp. 373-408. Original available here.
Summary of the above paper and thoughts on how Merlo and
Stevenson's method can be applied to Japanese:
- Study presentation held at June 12, 2002.
Discussed: Chapter 5, 5-1 to 5-3 of Takahashi, I., Kobayashi,
R. and Koyanagi, Y. 1992. Tookei Kaiseki (Statistical
Analysis), pp. 156-176. Baifuukan, Tokyo.
- June 26, 2002. Added preliminary material for a
talk to be held on July 10, 2002.
Paper discussed: Gildea, D. and Jurafsky, D. 2001. Automatic
Labeling of Semantic Roles. To appear in Computational
Linguistics. Original available
Summary of the above paper and thoughts on how it relates to Merlo
and Stevenson's method:
- Study presentation held at July 22 and July 24,
2002 during the Nakagawa/Tanaka laboratories Summer Seminar
Discussed: Chapter 2 of Watts, Duncan J. 1999. Small
Worlds: The Dynamics of Networks between Order and
Randomness. Princeton University Press.
- Research presentation held at August 9, 2002.
Discussed: Schulte im Walde, Sabine and Brew, Chris. 2002. Inducing
German Semantic Verb Classes from Purely Syntactic
Subcategorisation Information. Proceedings of the 40th Annual
Meeting of the Association for Computational Linguistics, July
7-12, 2002. Original available here.
Some useful links
Nakagawa Lab Study Session Portal Site
- Probably only interesting for lab-members (Japanese).
Hiroshi Nakagawa's homepage
- The homepage of my present supervisor.
- Makoto Kanazawa's
- The homepage of my former supervisor.
Manning's home page
- One of the authors of the famous Foundations of
Statistical Natural Language Processing.
- Beth C. Levin's home
- Wrote some very interesting books on the semantics-syntax interface
General academic links
- NEC Research Institute
- Useful for finding the academic papers of your interest.
- An Introductory Statistics Book and Online Tutorial for Help in
Statistical Analysis and Datamining
- A site on the above topics (Japanese).
Parsers for Japanese
- A parser developed by Tsujii Laboratory, Tokyo University.
Morphological and Syntactic LR Parser
- A parser from the Tokyo Institute of Technology.