第327回研究講演会開催案内

日時：2006年11月24日（金）13：30〜15：30

場所：山形大学工学部 ７号館203教室
         (〒992−8510　山形県米沢市城南4-3-16)

講師と演題:
１．講師：Andre Wlodarczyk　(Prof. of Charles de Gaulle University
(Lille 3) and Universite Paris Sorbonne (Paris 4))
   Currently : Visiting Researcher at the Nation Institute for Japanese
Language (Tokyo)
　　演題：Concept Reconstruction for Ontology-based Semantics
　　内容：If we want to reach better results in the field of semantic
analysis of linguistic phenomena certain foundational concepts (notions)
currently in use must be formally reconstructed. From the linguistic
(more generally semiological) point of view, semantic categories
(contents) must not be considered in separation from signs (units
defined originally as pairs of Form and Content).
However, the present approach is based on the assumption that meaning of
human-made signs as such being inaccessible for inspection, the only
reasonable solution for semantic research is modelling. Consequently, in
the described model,
(1) the sign is defined as a structure having a set of types of usages
named semions (defined as pairs of Signifier and Signified) as its
domain and a set of monadic infons (descriptions defined as pairs of
notations and a conjunction of true formulae),
(2) the structure of signs is characteristic of semiological Systems,
using signs is an instantiation of semiological Processes known
altogether as communication,
(3) the projection from Semions of a given Sign to Infons within the
System is known as assignment of descriptions (attributes) to the types
of usages of signs, whereas the projection from Signs to Infons within
the Process is said to be signification,
Ontologies are motivations (hierarchically structured foundations) of
semantic properties of signs. Semantics of natural languages is most of
all application domain specific. However, I claim that it is possible to
build meta-ontological (universal) hierarchies of concepts motivating
particular semantic solutions.
It will be shown how to build ontologies in order to describe linguistic
meanings using KDD algorithms integrated in "Semana" platform.

２．講師：Helene Wlodarczyk(Prof. of Universite Paris Sorbonne (Paris　4)):
   Currently : Visiting Researcher at Waseda University (Tokyo)
　　演題：Computer-aided Acquisition of Semantic Knowledge
　　　　　--the Category of Aspect
　　内容：We adopted the notation of semantic feature structures (simple
trees only) as a meta-language for describing Aspect in various
languages regardless of linguistic levels (morphological, syntactical,
lexical etc). We propose to describe the meaning of the Aspect category
as a pair of feature bundles : Analysis and Control. The Analysis of a
situation (viewed as a whole or as one of its moments or stages) is
considered as its endocentric aspect, and the Control of a situation
(viewed as a set of operations such as iteration, flow and intensity
modifications, composition) is defined as its exocentric aspect because
it is imposed from outside. These aspectual features occur and combine
diversely depending on the semantic type of the situation to which a
verb is related (due to hyperonymy/hyponymy lattice).
With the possibility to use the techniques of knowledge discovery in
databases (KDD) provided that the latter contain meta-linguistic
information, we present our theory of the Category of Aspect as the
first attempt of applying computational approximation-based methods in
order to determine the relevance and relative importance of the
parameters found. Only such detailed work with databases may be supposed
to allow us to offer formal, verified (i.e. experimentally tested) and
comparable cross-language definitions of semantic categories.
At the present stage of our research, the list of structures (in which
combinations of semantic parameters of Aspect can be observed)
describing the different usages of simple utterances is not exhaustive
as yet. However, while testing the coherence of our descriptions, we
could already improve our theory of Aspect.

　　問い合わせ先：
　　　後藤　源助（横山　晶一）
　　　山形大学工学部情報科学科
　　　 〒992−8510　山形県米沢市城南4-3-16
　　　 Tel: 0238-26-3275(3336)
　　　 E-mail: