In this communication, we advocate the use of upper level ontologies such as the Basic Formal Ontology (BFO) to enhance terminological resources and research. First, we present common issues in ontologized terminological work. Then, we review two projects that illustrate the potential advantages of integrating rigorous formal upper level ontologies. Finally, we discuss possible challenges and conclude with a summary of the benefits that such ontologies can bring to both terminological theory and practice.
Terminologies encode lexical and background knowledge that experts have about their domain of expertise. These resources can be associated with a more explicit ontology-like representation of the entities in the relevant domain. Such representations may include, for example, nonlexicalized concepts. This extends mere terminologies to more sophisticated knowledge representations. Being language independent, ontologized terminologies have the advantage of integrating multilingual terminologies. When augmented with axioms, they can be used in reasoning systems.
Terminological works, where they refer to
ontologies at all, generally use Gruber’s definition of
an ontology as “an explicit specification of a
conceptualization.”
However, a multiplication of ontological
metalanguages (categories and relations) tends to
create knowledge silos
In this communication, we present and discuss existing works integrating upper level ontologies, and underline the main advantages of augmenting terminological knowledge with categories and relations from an upper level ontology such as BFO. 2
Lack of rigorously defined categories and
relations. The interpretation of the
metalanguage is left to our intuitive understanding of
the terms used for expressing the used
categories and relations.
is a overloading
Multiplication of domain-specific, sometimes ad hoc, categories and relations.
When upper level categories are used,
limitation to a few top-most categories, which are
completed with domain-specific ones
The above limitations result in practical and research-related consequences for terminological works, which can be summarized as follows: Confusing and incompatible representations of the same domain.
Non-interoperable terminologies, which hinders the possibility of sharing and reusing terminological resources.
Non-generalizable observations of terminological phenomena, which hinders research towards a proper understanding of contentrelated principles governing term formation, definition composition, and conceptual system organization. This eventually hinders the development of widely (re)usable terminological tools, for example, for creating new terms and writing definitions.
Non-comparable results of terminological research for lack of a common well-defined domain- and language-independent metalanguage, which hinders the development of a mature integrated science.
These shortcomings can be addressed by adopting well-defined domain- and languageindependent upper level categories and relations (ontological metalanguage) of the sort accounted for in formal upper level ontologies. 180 3
A formal upper level ontology can be defined as “a representation of the categories of objects and of the relationships within and amongst categories that are to be found in any domain of reality whatsoever.” (Spear, 2006)
To illustrate the potential advantages for
terminology of using formal upper level ontologies, we
describe two projects that integrate such
ontologies. There are a few upper level ontologies that
can be used by mid-level or domain-specific
ontologies to define and relate their categories in a
non-ambiguous manner, using logical axioms if
needed. The projects described hereafter use,
respectively, the Descriptive Ontology for Linguistic
and Cognitive Engineering (DOLCE)
The KYOTO project aims at representing
domainspecific terms in a computer-tractable axiomatized
formalism to allow machines to reason over texts
in natural language
In Vossen et al. (2013), the authors extracted statements from texts about the Chesapeake Bay using Kybots, scripts based on ontological and linguistic patterns in annotated text. The results of baseline fact extraction were compared with Kybot extraction and Cterm extraction, both of which utilize the KYOTO ontology. The result was that the baseline and Kybot profiles had high recall, 100% and 91% respectively. The baseline had low precision (18%), whereas the precision of the Kybot profiles was better, though not optimal, at 31%. In short, leveraging ontological information in domain-specific fact extraction NLP resulted in high recall and improved precision. 3.2
The second project consists in analyzing the
contents of definitions using the categories and
relations of BFO
The results of the pilot study reported in Seppa¨la¨ (2012; 2015b) show that these BFOTemplates account for about 75% of the contents of definitions of terms from 15 distinct domains. The rest of the definition contents can be described using the BFO categories and relations.
The well-defined BFO vocabulary can thus be used as a metalanguage to describe definition contents, term formation, and the organization of conceptual systems in a way that research findings can be compared and integrated. In practice, BFO-based ontologized terminologies would have the advantage of being interoperable, as it is already the case for the mid-level and domainspecific ontologies (and the corresponding terminologies) that extend BFO, such as the Ontology for Biomedical Investigations (OBI) and the Ontology for Biobanking (OBIB)1. 4
Using upper level ontologies may sometimes prove challenging. Possible issues may be: Upper level ontologies evolve and their categories are, at times, still under development. 1For a full list, see http://ifomis. uni-saarland.de/bfo/users. For an illustration of interoperability and its advantages, see the presentation on The OBIB Ontology for Biobanking, by Chris Stoeckert, Jie Zheng, and Mathias Brochhausen http://ncorwiki.buffalo.edu/index.php/ CTS_Ontology_Workshop_2015.
In those cases, it may not be straightforward under which category to place a term.
Specifications of the upper level ontology may be sparse and lacking, and sometimes too formal (OWL, first order logic) to be easily understood by terminologists.
An adequate use requires familiarity with the upper level ontology chosen.
A solution to such issues would be to use
existing mappings of WordNet to upper level
ontologies as aids for integration. A future mapping
of WordNet to BFO should facilitate the
integration of BFO in terminological projects
We saw that ontologized terminologies present a number of shortcomings that can be addressed by integrating a formal upper level ontology. We illustrated the advantages of such an enhancement by reviewing two projects that use such ontologies. To summarize, the main benefits of using a language- and domain-independent upper level ontology are, on the practical side, the possibility to integrate multilingual and multi-domain terminological resources with one another and with information system tools. The latter can thus use the inferences drawn on the basis of the upper level ontology to reason over and manipulate multilingual natural language texts. Using a well-defined formal upper level ontology as a basis for terminological work would make sharing and reuse of terminologies easier: identifying and sharing common terms, constructing new definitions using the same building blocks (information types and logical axioms), etc. Such a framework avoids semantic conflicts and need for mapping.
On the research side, using a well-defined ontological metalanguage allows: carrying out rigorous and comparable conceptual analysis work in terminology; making language- and domainindependent generalizations about term formation, definition content structure, and terminological systems’ organization, which can help develop empirically based content standards and writing aid tools; creating comparable research results that contribute to developing a mature integrated terminological science.
Moreover, a metalanguage using the categories and relations of an upper level ontology for describing terminological data (for example, terms’, definitions’, and conceptual systems’ structure) can fruitfully complement any terminological resource whether or not already ontologized. Cimiano et al. (2011) propose, for example, a model to formally link lexicons (with relevant linguistic descriptions) to ontologies.
Using more specifically a BFO-based metalanguage would further enhance our understanding of the relationship between the lexical, linguistic, conceptual, and ontological levels of terminologies. Indeed, BFO is a realist ontology that represents the things that exist in the world and the relations between them, independently of our conceptualizations thereof. A BFO-based metalanguage may thus provide an additional level of understanding to existing descriptive frameworks.
We therefore encourage terminologists to fully embrace the best ontological practices to enhance their research and resources.
This work was supported in part by the Swiss National Science Foundation (SNSF) and by the NIH/NCATS Clinical and Translational Science Awards to the University of Florida UL1 TR000064. The content is solely the responsibility of the authors and does not necessarily represent the official views of the SNSF, the National Institutes of Health, or the NCTE. Thanks also to Aure´lie Picton and Barry Smith.