November
10 2003 - 9 a.m. to 6.30 p.m. followed by dinner
Hotel
Prindsen -
Program
9 Arrival.
Welcome and brief presentations of participants.
Coffee. Tea
Presentations I
9.30 Per
Anker Jensen & Jørgen Fischer Nilsson
“Towards
an Ontology-Based Semantics for Prepositions”
10 Discussion
10.15 Lone
Bo Sisseck
“Semantic relations between
concepts in Danish domain specific texts”
10.45 Discussion
11.00 Break. Juice. Fruit
11.15 Ekaterina
Mhaanna
“Terminology of Terminology: Standardization of
Concept Relations - Problems and Options”
11.45 Discussion
12.00 Lunch
followed by guided tour to the
http://www.vikingeskibsmuseet.dk./
Presentations II
14.30 - 16 Barbara
Partee & Vladimir Borschev
“For a Fire, To Be is To Burn: Ontology and
Semantic Bleaching”
Topic: The relation between ontology and the choice of “bleached verbs”
used in the Russian Genitive of Negation construction. This will focus on a
very specific part of our Genitive of Negation work and relate it to the
concerns of OntoQuery. (See abstract below)
16 Break.
Coffee. Tea. Cake
Presentations III
16.15 Bodil Nistrup Madsen, Hanne Erdman Thomsen
& Carl Vikner
“Principles
of an IT-System for Concept Analysis”
16.45 Discussion
17 Troels
Andreasen, Henrik Bulskov & Rasmus Knappe
“The
OntoQuery Prototype”
18 Discussion
and wrap-up
Dinner at
restaurant Snekken
http://www.snekken.dk/restaurant.html
Towards
an Ontology-Based Semantics for Prepositions
This paper addresses the
elaboration of a relation-logical compositional semantics for the meaning
content of nominals using formal ontologies as semantic domains. Prepositions
are conceived as representing binary relations between concepts in the
ontology. The ontology comes with ontological affinities specifying the admissible
ontological relations for ruling out ontological mismatches. The paper focuses
on the semantics of prepositional phrases examining in particular
disambiguation of phrases with multiple prepositional phrases utilizing the
ontological affinities. Danish, which
offers a rich system of prepositions, is used in the example material
accompanied by English translations.
My
project aims at identifying and extracting conceptual relations in Danish
domain specific texts in order to be able to generate ontologies automatically.
It is a very time-consuming task to build an ontology manually. It will
therefore be of great interest to be able to construct ontologies
automatically. Earlier and ongoing projects throughout the world have already
developed various methods of more or less automatic term and relation
extraction from both a statistical and a linguistic approach. However, there have hardly been any attempts
to investigate Danish texts in order to find linguistic patterns that could
indicate relations between concepts and eventually serve as a tool to build
ontologies automatically.
My project is related to the OntoQuery (OQ) project, which makes use of
ontologies in order to search for information in text databases. For further
details please see: www.ontoquery.dk. In the (manually build) OQ ontology, which is
based on texts from the nutrition domain, only the generic is-a relation has been used. My first aim is to analyze the same
corpus manually in order to locate linguistic patterns that indicate generic
relations between concepts. The results will be compared with the existing OQ
ontology. I will present the results of this analysis, which will serve as
foundation for my future work with identifying and automatically extracting
relations between concepts in a new arbitrary domain.
NB!
For a Fire, To Be is
To Burn: Ontology and Semantic Bleaching
The
talk concerns the relation between ontology and the choice of “bleached verbs”
used in the Russian Genitive of Negation construction. This will focus on a
very specific part of our Gen Neg work and relate it to the concerns of
OntoQuery. We follow Babby (1980) in analyzing the intransitive Gen Neg
construction as limited to existential sentences; Babby noted that the class of
verbs permitted in Gen Neg sentences is constrained but open. Lexical verbs
other than byt’ ‘be’ and suščestvovat’ ‘exist’ used in Gen
Neg sentences are often characterized as “semantically empty” or “bleached”. We
give an account of semantic bleaching that starts from the assumption that each
verb used with Gen Neg still retains its normal meaning, and that the
“bleaching” involves the interaction of the normal verb semantics with a
constructional presupposition of the sort illustrated in our title slogan, “For
a fire, to be is to burn.” The verb ‘burn’ would not function as an existential
verb with a noun like book or house, but can with the noun fire. The factors that contribute to
supporting “local” equivalences like “to be is to burn” (true of fires but not
true in general) include lexical information, general encyclopedic or common
sense information, and specific contextually salient information. We discuss
the dynamic interplay of speaker and hearer in encoding and decoding covert
assumptions contributing to such “local” equivalences.
We
are working on a project called CAOS - Computer Aided Ontology Structuring - whose
aim is to develop a computer system designed to enable semi-automatic
construction of ontologies, or concept systems. CAOS supports terminological
concept modelling. The backbone of this concept modelling is constituted by
characteristics modelled by formal feature specifications, i.e. attribute-value
pairs. The use of feature specifications is subject to a number of principles
and constraints such as, for instance:
· function from attributes to values
In this paper we would like to discuss some of these principles and to show why they are necessary and how they contribute to determine the structuring of the ontologies in CAOS and to facilitate the work of the ontologist.
In this presentation we will touch upon the approach taken in the
OntoQuery project to query evaluation. We will give a short demo of the latest
prototype and with this as a starting point discuss generation and comparison
of "concepts", in the form of expressions in the
description language Ontolog.
With respect to generation we will introduce to the prototype principle
of unambiguous tagging + simple NP recognition and discuss currently considered
alternatives. The key point in comparison is the similarity exploited in query
evaluation. We will discuss how to measure and apply similarity in a
content-based information retrieval
environment, where text is indexed by descriptions (sets of concepts), which
refer to the ontology.