Pamela Faber1, Ricardo Mairal2, Pedro Magaña3
1 University of Granada - Buensuceso, 11 18002 Granada, Spain - pfaber@ugr.es
2 UNED (Madrid) - Senda del Rey, 7 28040 Madrid, Spain - rmairal@flog.uned.es
3 CEAMA (Centro Andaluz de Medio Ambiente) - Av. del Mediterráneo s/n 18006 Granada, España - pmagana@ugr.es
Abstract
Ontologies have been criticized because they are not suffi-
ciently flexible, and thus cannot capture the dynamism and
complexity of reality. However, they have increasingly
come into focus because of the need for knowledge man-
agement in both general and specialized knowledge do-
mains. EcoLexicon is a frame-based visual thesaurus on the
environment that is gradually evolving towards the status of
a formal ontology. For this purpose, the information in its
relational database is in the process of being linked to the
ontological system of FunGramKB, a multipurpose know-
ledge base that has been specifically designed for natural
language understanding with modules for lexical, grammat-
ical, and conceptual knowledge. This enables the explicita-
tion of specialized knowledge as an extension of general
knowledge through its representation in the domain-specific
satellite ontology of a main general ontology.
1. Introduction
A domain-specific ontology of concepts within a certain
field, along with their relations and properties, is a new
medium for the storage and propagation of specialized
knowledge (Hsieh et al. 2010). Ontologies reflect a particu-
lar conceptualization of reality through the explicit defini-
tion of concepts (terms), representing domain entities, and
relations. From a more linguistic perspective, Sowa (2000:
492) defines an ontology as "a catalogue of the type of
things that are assumed to exist in a domain of interest D,
from the perspective of a person who uses a language L for
the purpose of talking about D." In this respect, one way to
enrich ontology elements is by including linguistic infor-
mation and structure (Buitelaar et al. 2009). In this sense, a
multilingual terminological knowledge base is a valuable
knowledge resource since it is composed of signs in vari-
ous languages that designate concepts, corresponding to
mental representations of phenomena in the real world.
Copyright © 2011, Association for the Advancement of Artificial
Intelligence (www.aaai.org). All rights reserved.
Multilingual information is in great demand today by insti-
tutions (Montiel-Ponsoda et al. 2010), and multilingualism
in ontologies benefits society because it helps to reduce
confusion regarding conceptual reference in international
communication. This has evident implications for e-
learning and knowledge acquisition.
EcoLexicon (http://ecolexicon.ugr.es) (Faber et al. 2006;
Faber et al. 2007) is a multilingual visual thesaurus on the
environment in English, Spanish, and German (currently
under expansion to French, Russian, and Modern Greek).
Its purpose is knowledge acquisition with a view to spe-
cialized text generation by users such as scientific writers,
translators, and environmentally-aware sectors of the gen-
eral public. The conceptualization process of the resource
first involved the semi-automatic extraction of concepts
and relations from domain-specific documents (Eriksson
2007), based on semantic patterns and lexical markers.
Other environmental resources used for information extrac-
tion were general upper-level ontologies, such as SIMPLE
(Lenci 2000) and environmental ontologies such as SWEET
(Raskin and Pan 2005).
The next step was the manual creation of an ontology
‘scaffold’ made up of basic class hierarchies and relations.
The ontology itself is organized around direct representa-
tions of physical objects and processes (e.g. alluvial fan,
erosion, weathering, etc). This basic set of concepts act as
a scaffold, and their natural language descriptions provide
the semantic foundation for data querying, integration and
inferencing (Samwald et al. 2010). This scaffold could not
be generated automatically since some of the structures and
entity labels in the database needed to be changed and re-
interpreted so that there would be agreement between con-
ceptualizations in different languages. Concepts and rela-
tionships are currently in the form of semantic networks or
concept maps implemented with ThinkMap. Concept maps
have been successfully used in other domains, such as
nutrigenomics (García Castro et al. 2006).
Linking a Domain-Specific Ontology
to a General Ontology
564

In EcoLexicon, environmental concepts are codified in
terms of natural language definitions that are visually rep-
resented as a network of both hierarchical and non-
hierarchical semantic relations extracted from a multilin-
gual corpus of specialized texts. This is in consonance with
cognitive semantics (Talmy 2000), which claims that lexi-
cal meaning is a manifestation of conceptual structure. The
meaning of words thus does not depend on the world itself,
but rather on our categorization of the world (Evans Ber-
gen and Zinken 2007). Similarly, Gahegan et al. (2008)
affirm that the concepts and relations used to describe the
world are constructed by humans. In the vision of concep-
tual organization offered by cognitive semantics, lexical
items are regarded as conceptual categories of distinct yet
related meanings that exhibit typicality effects.
Even though this representation still needs to be further
enriched and systematized so as to allow more sophisti-
cated reasoning processes, it permits EcoLexicon to be
connected to other ontologies and resources. Accordingly,
this paper describes the integration of EcoLexicon into
FunGramKB, a multipurpose general knowledge base,
which uses COREL as a representation language. It explains
how the general concepts in FunGramKB can be extended
and reused in deep semantic representations in a domain-
specific ontology.
This paper is organized as follows Sections 2 and 3 de-
scribe the conceptual organization and design of EcoLexi-
con and FunGramKB, respectively. Section 4 explains the
integration process, the semantic representation as applied
to specialized knowledge, the promotion/demotion of gen-
eral language concepts in the domain-specific ontology,
and the advantages gained from contextualizing a special-
ized knowledge resource within a general ontology. Sec-
tion 5 presents the conclusions derived, and Section 6 lists
the references cited.
2. EcoLexicon
EcoLexicon is a specialized knowledge resource on the
environment. It is hosted in a relational database, which is
currently being linked to an ontology for reasoning tech-
niques and user queries (Leon, Magaña, and Faber 2008;
León and Magaña 2010). As such, it focuses on: (1) con-
ceptual organization; (2) the multidimensional and multi-
lingual nature of terminological units; (3) the extraction of
semantic and syntactic information through the use of
multilingual corpora. Based on cognitive semantics and
situated cognition (Barsalou 2008), the information stored
in EcoLexicon is structured in terms of propositions and
knowledge frames (Fillmore 1985) that are organized in an
ontological structure, which focuses on perceptual infor-
mation and semantic relations.
EcoLexicon can be regarded as a linguistically-based on-
tology since its conceptual design is derived from informa-
tion semi-automatically extracted from specialized texts
and the structure of terminological definitions. Its top-level
concepts are object, event, and attribute categories. In
EcoLexicon, abstract concepts include theories, equations,
and units for measuring physical entities. In contrast, phys-
ical or concrete concepts are those occupying space and
occurring over a period of time. They include natural enti-
ties, geographic landforms, water bodies, constructions,
and the natural and artificial process events in which they
can potentially participate.
In Ecolexicon, the most generic or top-level categories
of a domain are configured in a prototypical domain event
or action-environment interface (Barsalou 2008), called the
Environmental Event (EE) (Faber, Márquez, and Vega
2005) (see Figure 1).
Figure 1. Environmental Event (EE)
The EE has two types of agent that can initiate
processes. Such agents can be inanimate (natural forces) or
animate (human beings). Natural agents, such as water
movement (e.g. waves, tides, and currents) and atmospher-
ic phenomena (e.g. winds and storms) cause natural
processes such as littoral drift and erosion in a geographic
area such as the coast. These processes affect other entities
or PATIENTS (e.g. beaches, sea ports, and seabed) which as
a RESULT, may suffer changes (e.g. loss/deterioration/
creation of beaches, and modifications in seabed composi-
tion). HUMAN AGENTS can also implement ARTIFICIAL
PROCESSES (e.g. constructions), which can generate or
prevent EFFECTS normally caused by natural processes.
For instance, a TSUNAMI, as a large high-velocity wave,
can initiate a FLOOD-EVENT, which affects a patient
(LANDFORM or LAND AREA) and produces a certain result
(EROSION, MODIFIED LANDFORM. etc.). Alternatively, with-
in the context of other processes or events, it can be re-
garded as the result of the displacement of the sea floor
(i.e. sudden faulting, landsliding, or volcanic activity). This
565