Conference PaperPDF Available

Abstract and Figures

Understanding the differences underlying the scope, usage and content of language data requires the provision of a clarifying terminological basis which is integrated in the metadata describing a particular language resource. While terminological resources such as the SIL Glossary of Linguistic Terms, ISOcat or the GOLD ontology provide a considerable amount of linguistic terms, their practical usage is limited to a look up of a defined term whose relation to other terms is unspecified or insufficient. Therefore, in this paper we propose an ontology for linguistic terminology, called OnLiT. It is a data model which can be used to represent linguistic terms and concepts in a semantically interrelated data structure and, thus, overcomes prevalent isolating definition-based term descriptions. OnLiT is based on the LiDo Glossary of Linguistic Terms and enables the creation of RDF datasets, that represent linguistic terms and their meanings within the whole or a subdomain of linguistics.
Content may be subject to copyright.
Christiani Lehmanni inedita, publicanda, publicata
OnLiT : An ontology for linguistic terminology
huius textus situs retis mundialis
dies manuscripti postremum modificati
occasio orationis habitae
Language, data and knowledge 2017. 19-20 June 2017 in
Galway, Ireland
volumen publicationem continens
Gracia, Jorge et al. (eds.), Language, data, and knowledge.
First International Conference, LDK 2017, Galway, Ireland,
June 19-20, 2017, Proceedings. Berlin & Heidelberg:
annus publicationis
OnLiT: An Ontology for Linguistic Terminology
Bettina Klimek1(B
2, Christian Lehmann3,
Christian Chiarcos4, and Sebastian Hellmann1
1InfAI, University of Leipzig, Leipzig, Germany
2Insight Centre for Data Analytics,
National University of Ireland Galway, Galway, Ireland
3University of Erfurt, Erfurt, Germany
4Applied Computational Linguistics,
Goethe-University Frankfurt, Frankfurt, Germany,,,
Abstract. Understanding the differences underlying the scope, usage
and content of language data requires the provision of a clarifying termi-
nological basis which is integrated in the metadata describing a partic-
ular language resource. While terminological resources such as the SIL
Glossary of Linguistic Terms, ISOcat or the GOLD ontology provide a
considerable amount of linguistic terms, their practical usage is limited to
a look up of a defined term whose relation to other terms is unspecified or
insufficient. Therefore, in this paper we propose an ontology for linguistic
terminology, called OnLiT. It is a data model which can be used to rep-
resent linguistic terms and concepts in a semantically interrelated data
structure and, thus, overcomes prevalent isolating definition-based term
descriptions. OnLiT is based on the LiDo Glossary of Linguistic Terms
and enables the creation of RDF datasets, that represent linguistic terms
and their meanings within the whole or a subdomain of linguistics.
Keywords: Linguistic terminology ·Linguistic linked data ·LiDo
1 Introduction
The research field of language data has evolved to encompass a multitude of inter-
disciplinary scientific areas that are all more or less closely bound to the central
studies of linguistics. Understanding the differences underlying the scope, usage
and content of language data provided by diciplines such as linguistics, computa-
tional linguistics, digital humanities or content analytics, requires the provision
of a clarifying terminological basis which is integrated in the metadata describ-
ing a particular language resource. Moreover, the comparative use of resources
Springer International Publishing AG 2017
J. Gracia et al. (Eds.): LDK 2017, LNAI 10318, pp. 42–57, 2017.
DOI: 10.1007/978-3-319-59888-8 4
OnLiT: An Ontology for Linguistic Terminology 43
of different languages presupposes that they use the same conceptual framework
and terminology. This demand for specifying linguistic terminology has been
addressed mainly by linguists in creating look-up resources such as books, e.g.
the lexicon of linguistics (Bußmann et al. 1996), online registries (e.g. ISOcat1
(Kemps-Snijders et al. 2009), the SIL Glossary of Linguistic Terms2(Loos et al.
2004) and the CLARIN Concept Registry (Schuurman et al. 2016) or Web pages
such as the online encyclopedia of linguistics3.
While all these resources provide a considerable amount of linguistic terms,
their practical usage is limited to a look up of a term whose relation to other
terms is unspecified or too general. In this respect the available data resources
of linguistic terminology fail to provide a meaningful representation of a linguis-
tic term leaving it isolated within the whole domain of linguistic terminology.
Retrieving more information about linguistic concepts necessitates reading their
definitions and looking up further words that are contained in it, which might be
also defined terms in the database. This procedure is not only time-consuming
and impractical but also results in implicit and vague specifications of linguistic
terms. This is the argument from the viewpoint of usability. However, mainte-
nance of a consistent conceptual-terminological framework likewise requires that
the relations among concepts be standardized and that, for each concept, the
relevant relations be specified. A set of isolated terms cannot be kept consistent.
In this paper we propose an ontology for linguistic terminology, called OnLiT,
as a data model which can be used to represent linguistic terms and concepts
in a semantically interrelated structure. Every terminological dataset evolving
from OnLiT will result in a data graph which is easy to navigate for human
users, machine-processable for semantic applications and will serve the purpose
of directly and indirectly interrelating linguistic terms and concepts throughout
the whole dataset. The OnLiT model is based on the Linguistic Documentation
(LiDo) database by Christian Lehmann4,5, who established a relational network
which represents linguistic terminology that defines and delimits a term by relat-
ing it to the linguistic concept it encodes and also by including a set of specify-
ing conceptual relations (Lehmann 1996). What is more, the proposed model is
independent of the particular language of the terms and thus allows integration
of terminological networks in different languages and multilingual terminologi-
cal networks. By transforming the structure of the LiDo relational database to
RDF, the OnLiT data model aims to provide the following contributions:
to enable a semantic search for linguistic terms and concepts,
to provide unique reusable and citable identifiers for each data entry,
4A browseable version of the database is available at: http://linguistik.
5Christian Lehmann is the data owner of LiDo and permitted to derive the OnLiT
data model from it.
44 B. Klimek et al.
– to enable the creation of conceptually consistent terminological datasets
that broadly interconnect and cover linguistic terms in a required linguis-
tic (sub)domain,
– to establish the possibility for extending the data model and enriching an
OnLiT dataset with external data,
to allow free and open reuse of the OnLiT data model.
The remainder of the paper is structured as follows. Section 2gives an
overview of relevant related work. Following an outline of the LiDo database as
basis for OnLiT in Sect. 3, the OnLiT data model is presented in Sect. 4.1. Fur-
ther, the purpose, domain and requirements of OnLiT are presented in Sect. 4.2
and the modelled concepts, terms and the established relations between them are
discussed in Sects. 4.3 and 4.4. Finally, in Section 5the paper concludes giving
a brief summary and a prospect of future work.
2 Related Work
An investigation of available datasets (excluding the LiDo database which is
presented in Sect. 3) that contain models of representing linguistic terminology,
resulted in two different types of data.
(i) Linguistic term bases that offer a term look-up via a Website: Resources
such as the aforementioned ISOcat registry or SIL Glossary of Linguistic Terms
(GLT) are mainly aimed at human users. Their underlying semantic structure is
rather flat providing definitions and very unspecific superordinate and subordi-
nate concept relations such as is a or has kinds. In the GLT, further, terms in a
term entry can be traced by the user via established links. Navigating through
ISOcat is harder since it provides a wide range of different “views” and “groups”
which provide linguistic terminology in general but also specify linguistic terms
in a specific language data model, e.g. the “STTS group” or “CLARIN group”.
In this regard such linguistic term bases have no underlying data model that
represents linguistic terminology in an interrelating holistic structure. What is
more, the arbitrary structure of the data models, which represent the linguistic
term entries in alphabetical order (as in GLT) or according to linguistic views or
linguistic data groups (as in ISOcat) is neither sufficient nor suitable for gaining
comprehensive knowledge about a linguistic term in the domain of linguistics.
A recent project, the CLARIN Concept Registry (Schuurman et al. 2016), has
taken over the work of ISOcat and promises to define terms in a stricter manner,
although still providing very limited structural and relational information.
(ii) Linguistic concepts represented as Linked Data ontology: In order to
enable the description of linguistic data, formalized ontological models emerged
within the realm of the Semantic Web. The most significant model for the sci-
entific description of human language is the General Ontology for Linguistic
Description (GOLD)6(Farrar and Langendoen 2003; Farrar 2010). It provides
a taxonomy of nearly 600 linguistic concepts, which have been constructed from
OnLiT: An Ontology for Linguistic Terminology 45
the GLT, and formalizes 83 relations (i.e. 76 object properties and 7 data prop-
erties). GOLD has been designed to support Community-of-Practice Extensions
(COPEs), meaning that it is a recommended upper model for ontologies of lin-
guistic terminology that can define their concepts as sub-concepts of GOLD
concepts (Farrar and Lewis 2007). This mechanism has been adopted by several
ontology providers, e.g., (Wilcock 2007; Good et al. 2005;Goeckeetal.2005). In
that usage and because the terms provided by the GLT have been transformed
into concepts in GOLD, linguistic terms and concepts are not distinguished any
more. The concepts are only defined within the domain of linguistic descrip-
tion but not in the more general domain of linguistics. In addition, the variety
of object properties assigned to the concepts are very specific and interrelate
mostly only two concepts, which leaves the majority of the concepts unrelated.
The established relations are either too specific or too general to derive the mean-
ing of a concept within the domain of linguistics, e.g. a “grapheme” concept is
defined within the taxonomy as a “FormUnit” concept, which is a “LinguisticU-
nit” concept, which is an “Abstract” concept. It has no further relations to other
concepts, e.g. to “Character”, which only implicitly states in its rdfs:comment
that it is “similar to grapheme”. Also, it is unclear why the “Character” con-
cept ist not also modelled as a subconcept of “FormUnit”. These are solvable
issues, however, the development of GOLD and the community process stopped
in 2010. Despite the wealth of linguistic concepts in GOLD it would be a very
inconcise model for linguistic terminology, due to the lack of terms relating to
the concepts and due to the complexity of relations which is aimed at a subfield
of descriptive linguistics but not at representing linguistic concepts in a more
encompassing scope of the domain of linguistics.
These two primary kinds of sources for a model of linguistic terminology
can be summarized as being either term-focussed or concept-focussed. A coher-
ent model of linguistic terminology, however, presupposes explicitly establishing
both linguistic concepts and terms and placing them into the whole domain of
linguistics. To conclude, to our knowledge there is - with the exception of the
LiDo database - no data model available that appropriately describes linguistic
terminology as the domain of linguistic terms that encode linguistic concepts
which are interrelated in a meaningful way.
3 The LiDo Glossary of Linguistic Terms as OnLiT
The LiDo database7as it is available in its current form as a browsable glossary
of linguistic terms has a thirty year old history. Christian Lehmann started to
collect and systematize his terminological knowledge as a general comparative
linguist by introducing a documentation system for linguistics in 1976 (Lehmann
1976). Twenty years later its technical implementation in 2006 resulted in the
7It has to be mentioned that LiDo encompasses also bibliographical data that is
referenced to the terms. This bibliographic part of the dataset is, however, not focus
of this paper and, hence, not further discussed.
46 B. Klimek et al.
LiDo Web frontend which is based on a relational database8that has been con-
tinuously updated and extended by Christian Lehmann ever since. To date, the
LiDo term and concept data encompasses more than 4500 unique linguistic con-
cepts and over 15000 terms, most of them in English, German, Spanish and
Portuguese. Moreover, each concept is interrelated to at least one other concept
which yields a coherent terminological data graph. Editing and curating this
considerable data size is enabled by a manageable set of relations which fulfill
the self-imposed requirement to explicitly express a direct relation between two
linguistic concepts (Lehmann 1996). This is achieved by the two formal rela-
tions of coordination and subordination which generate an overall taxonomic
and meronomic structure and 14 subrelations of those that permit a semanti-
cally specified interrelation of concepts. As a consequence, the data structure
underlying the LiDo term and concept data inheres the following criteria which
we see as essential for describing terminological data:
explicit representation of concepts and terms as separate resources,
meaningful interrelation of concept and term data,
an easy to use and editable data structure.
Therefore, the underlying LiDo data structure does not only permit an appro-
priate representation of the domain of linguistic terminology but also implicitly
contains an ontological modelling of the domain. These two aspects finally moti-
vate the reuse of the LiDo model as a data basis for creating OnLiT.
4 The Ontology for Linguistic Terminology
4.1 Components of the OnLiT Model
The OnLiT vocabulary is freely available under the URL http://lido. open for any kind of reuse under the CC BY
4.0 license10. As a Linked Data model which is based on the Web Ontology
Language (OWL11), OnLiT consists of a hierarchy of conceptual classes which
represent commonality among a variety of entities, i.e. the so-called instances,
individuals or resources of a dataset. The semantics of entities within the ontol-
ogy is formally defined by class usage restrictions that can hold between classes
and are encoded within relations. Relations are formally expressed as object
properties or as data properties.
An overview of the class modelling in OnLiT is given in Fig.1and a
detailed view of the object property structure is provided in Fig.2. For mod-
elling the domain of linguistic terminology, the OnLiT vocabulary contains only
8This database is used to render the LiDo Website but not publicly available. The
database was used in order to conduct the presented research.
9In case of unavailability:
OnLiT: An Ontology for Linguistic Terminology 47
Fig. 1. Class diagram of the OnLiT model.
7 classes: Concept,Term,Identifier (with ConceptID and TermID as sub-
classes), Abbreviation and Editor. Only the first two are essential and should
in any case contain instances (a more detailed presentation of their usage is
given in Sect. 4.3). An Concept instance describes a language-independent men-
tal entity which is encoded in different language-specific terms. As such concepts
are cognitively defined as substantial meanings which are realized by a linguistic
sign, which is then the term associated with the concept. In order to be able
to identify and refer to such a mental (as opposed to the formal understanding
of concepts as classes in OWL!) conceptual instance, it needs to be somehow
denominated with a humanly readable name. This can be done by an arbitrary
string identifier or by using the term expression that standardly encodes the
concept in some language as, i.e., there could be a ‘noun’ Concept instance and
anoun Term instance. The former, however, serves only as a conventionalized
naming method12 for a cognitive and language-independent meaning while the
latter is a linguistic expression of the English language. This distinction is similar
to the division of sense IDs which are associated to lexical entries in datasets such
as WordNet13.TheAbbreviation class is established, because linguistic terms
can have various conventional abbreviations assigned. This is common practice
in language description and might be, therefore, useful for some dataset creators.
Meta-information provided by the Identifier and Editor classes are added for
convenience, because they tend to be included in other dataset formats, such
as tables and relational databases. These can be directly used in case already
12 In the LiDo database Latin expressions are used to a large extent to denominate the
concept entries.
48 B. Klimek et al.
Fig. 2. Inheritance diagram of OnLiT object properties and subproperties.
existing datasets of linguistic terms in such formats shall be transferred into
RDF with the OnLiT model. However, more fine-grained Linked Data vocabu-
laries are available for representing the metadata of a dataset, e.g. DCMI terms14
or PROV-O15, which are easily integrable due to the interoperability of Linked
Data vocabularies.
With regard to the relations, there are three main object properties estab-
lished in OnLiT that interrelate instances of (1) terms with terms, (2) terms with
concepts and (3) concepts with concepts. The term-termRelation property can
be used to specify the relation between noun Term instances, on the one hand,
OnLiT: An Ontology for Linguistic Terminology 49
and adjective and verb Term instances, on the other, in a dataset. That way
adjective and verb terms can be included in a dataset if they are desired to
be described as linguistic terms (cf. Sect. 4.3 below) and related to their cor-
responding noun Term resources16 which are then interrelated to the respective
Concept instance they encode. A term-conceptRelation is established in order
to enable the assignment of the Term instance to its associated Concept instance.
The most structuring are the concept-conceptRelation object properties.
Because these are divided in the subproperties of coordinatingRelation and
subordinatingRelation they add to the taxonomic and meronomic structure of
the Concept data and therewith also of the Term data within an OnLiT dataset.
The twelve subordinatingRelation subproperties are intended to establish a
semantically more specific interrelation between concepts (a more detailed pre-
sentation of their usage is given in Sect.4.4).
Overall the OnLiT model is of manageable size but yet provides sufficient
explicitly modelled semantic interrelations to create a consistent dataset of lin-
guistic terminology.
4.2 Purpose, Domain and Requirements of OnLiT
There are two main purposes pursued by the OnLiT model. First, it serves as
the conceptual foundation for an RDF dataset of the LiDo Glossary of Lin-
guistic Terms including the whole relational database of its Term and Concept
data. Second, it provides users and creators of language data in general as well
as the community of Linguistic Linked Open Data in particular with a means
to easily set up and/or semantically interconnect various linguistic terminolog-
ical datasets. Moreover, its basic properties are transferable to the definition of
terminological datasets for other scientific disciplines.
The domain of linguistic terminology as represented by OnLiT is not
restricted to a certain definition of term. Thus, any expression that needs to
be described with OnLiT for theoretical or practical reasons can constitute a
Term or Concept resource in an OnLiT dataset. As a consequence, even proper
nouns denoting persons, e.g. Noam Chomsky or linguistically significant words of
a language, e.g. the grammatical verb be can be entries in an OnLiT term base.
In that respect Term entries in an OnLiT dataset are not limited to a narrow
definition of scientific term as being a common noun (Kamlah and Lorenzen
1967). Rather, this definition is broadened to allow individuals’ names, plain
lexemes or even adjectives and verbs to be included as terminological entries.
Given that OnLiT is based on the LiDo Glossary of Linguistic Terms, it meets
the same criteria as outlined in Sect. 3. In addition to that and in contrast to the
Lido data model, OnLiT is based on Semantic Web modelling principles. Due to
that, OnLiT based datasets fulfill the requirements of semantic and structural
interoperability which enable an easy reuse of data and further enrichment via
interlinking to external data sources.
16 This allows, for instance, to integrate the Term entries homonymous and govern and
relate them to homonymy and government.
50 B. Klimek et al.
We assume that datasets evolving from the OnLiT model will add to the cre-
ation of a comprehensive terminological knowledge graph of the field of linguistics
ranging from general and traditional linguistic terminology to the representation
of newly evolving or very specifically used terms and concepts.
4.3 Linguistic Concepts and Terms
The Term and Concept classes constitute the essential classes of an OnLiT
dataset since these contain the concept and term resources respectively. Two
relations can be specified between them, which express that a Term resource is
a standard or a non-standard term for a given concept. Their interrelations are
illustrated in Fig. 3, which exemplifies the triples for the Term instance noun and
the Concept instance ‘nomen substantivum’. Concept resources are unique, since
they are mental objects which are designated by a linguistic expression, i.e. the
Term resource. As a consequence, there can be multiple Term resources related to
a single Concept resource. Thus, there is also a Substantiv resource stated to be
the standard German term and also a Nennwort resource to be a non-standard
term for the Concept resource ‘nomen substantivum’. This can be achieved by
forming triples between Term and Concept resources via the two object prop-
erties stdTerm and non-stdTerm. This is the way of dealing with synonymous
terms. For a homonymous term, the relation to one Concept resource is selected
as stdTerm, and all the others are non-stdTerm.EachTerm resource can use the
property stdTerm for only one Concept resource, while it can be non-stdTerm
for more Concept resources. For instance, the German term Nomen is standard
for the concept ‘nomen’ and non-standard for the concept ‘nomen substantivum’.
Further, every Term resource should be explicitly assigned to a language. For that
purpose the language identifiers of the lexvo vocabulary17 are reused, because
they provide a precise language assignment as well as machine-readability.
The Concept resources can be further specified for additional information
by describing the definition, delimitation and history, analytic procedure, phe-
nomenology and example(s) via the respective datatype properties (cf. Fig. 3).
This information is provided by plain text and constitutes information a linguist
might have documented about a certain linguistic concept and which should be
included in the database. In fact, definition and examples are frequently found in
terminological datasets (e.g. in GLT or ISOcat) and can be simply transferred to
an OnLiT dataset by using these datatype properties. Even though information
stated in such plain text literals is not directly machine-readable and, therefore,
also not semantically explicit enough for automatic data processing, it is from
a human data consumer perspective very insightful. Eventually, the definitions
constitute indeed a useful information source that reveals information about a
concept, that can be formally modelled. The definition of the ‘nomen substan-
tivum’ Concept resource states that it is “a [. . . ] part of speech”, which can be
formalized via a subordinating relation between the given Concept resource and
another ‘part of speech’ Concept resource (this will be demonstrated in Sect. 4.4).
OnLiT: An Ontology for Linguistic Terminology 51
Fig. 3. Example modelling of the English Term instance noun and the Concept instance ‘nomen substantivum’ with OnLiT. (The
exemplary data for both resources used here and in other parts of this paper can be consulted under the English “noun” entry in the
LiDo database Web frontend:
52 B. Klimek et al.
Hence, textual information about linguistic concepts are not only most prevalent
in already existing terminological datasets, but also assist the OnLiT dataset cre-
ators in formally expressing their explicit defining relations to other concepts.
Conversely, a good definition incorporates the conceptual relations specified for
the concept.
To summarize, the representation of linguistic concepts and terms adheres
to the requirement of providing separate resources for both. What is more, the
relation that holds between a term and concept is modelled in OnLiT as a one
to one correspondence between a Term instance (having a single unambiguous
meaning) and the corresponding Concept instance (being the mental object of
that single meaning) it designates. This ensures a disambiguated traceability
and clarification of linguistic terms within the domain of linguistics. Also, the
OnLiT model provides a manageable but significant set of object and datatype
properties which specify Concept and Term resources in more detail and which
can be easily extended with further properties if need be.
4.4 Interrelating Linguistic Concepts
As presented in the previous section, there are only two object properties that
relate Term resources to Concept resources. The majority of relations is specified
in object properties which are established between two Concept resources. While
these relations could theoretically also hold between Term resources, this is not
done for a practical reason. Because multiple Term instances can refer to the
same Concept instance it is more economic to assign specific interrelations once
to the Concept instance, instead of repeating them on every Term instance that is
associated with the same Concept instance. This holds a fortiori for translations
of the terminological dataset into other languages. As a result, the semantic
specification is directly attached to the Concept resources and, therefore, also
indirectly to the Term resources via the term-conceptRelation subproperties
(as describ ed in the previous Section). Figure 4exemplifies how multiple Term
resources can encode a single Concept resource, which provides further semantic
specification through the concept-conceptRelation subproperties.
As is shown in Fig. 2, the 14 object properties which are at the lowest level of
the object property hierarchy are the most specific ones. In order to create a more
general taxonomic structure, these are systematized according to the superprop-
erties coordinatingRelation and subordinatingRelation. As a result, more
statements can be inferred that relate Concept instances on a broader seman-
tic level. Such inferred triples are expressed in Fig.4via the dashed arrows
connecting the Concept instances. There are two subproperties which yield a
coordinating relation and which are described as follows:
x isCross-RelatedWith y: States that a concept is somehow cross-related
with another concept, although the two are not sisters subordinate to a third
OnLiT: An Ontology for Linguistic Terminology 53
Fig. 4. Example modelling of the Concept instance ‘nomen substantivum’ with its interrelations to other concepts.
54 B. Klimek et al.
Example 1. nomen adjectivum (adjective)18 isCross-RelatedWith attributum
x contrastsMinimallyWith y: States that a concept contrasts minimally
with another concept.
Example 2. aspectus perfectivus (perfective)contrastsMinimallyWith aspec-
tus imperfectivus (imperfective).
The coordinatingRelation subproperties are symmetric properties, that
group semantically similar Concept instances by cross-referencing.
For creating subordinating relations between Concept instances twelve sub-
properties can be used:
x isAKindOf y: Is the most general subordinating relation, that states that
a concept is a kind of another superordinating concept. The interrelation of
concepts with this property creates a taxonomy.
Example 3. linguistica (linguistics)isAKindOf scientia rerum humanarum
(human science)isAKindOf scientia (science) isAKindOf activitas (activity).
x asAClassIsA y: States that if a concept x is taken to represent a class,
this is a subclass of another class concept.
Example 4. nomen adjectivum (adjective)asAClassIsA pars orationis (word
x isAClassOf y: States that a concept represents a class.
Example 5. pars orationis (word class)isAClassOf dictio (word).
x isElementOfTheRelation y: States that a concept is an element of a
relation represented by another concept.
Example 6. allomorphum (allomorph)isElementOfTheRelation allomorphia
(allomorphy ).
x isOperatorOf y: States that a concept is an operator of an operation
represented by another concept.
Example 7. affixum (affix)isOperatorOf affixio (affixation).
x isPartOf y: States that an entity falling under concept x is a part of an
entity falling under another concept. Concepts that are interrelated with this
“part-whole” property will create a meronymy.
Example 8. casus (cas e )isPartOf declinatio (declension).
declinatio (declension)isPartOf flexio (inflection).
flexio (inflection)isPartOf systema morphologicum (morphology).
systema morphologicum (morphology)isPartOf systema grammaticum
systema grammaticum (grammar)isPartOf systema linguae historicae (lan-
guage system).
18 For better comprehensibility the standard English Term instances corresponding to
the given Concept instancesaregiveninbrackets.
OnLiT: An Ontology for Linguistic Terminology 55
x isProperty-AspectOf y: States that a concept represents a characteristic
or possible aspect or property of its superordinate concept.
Example 9. arbitrarietas signi (arbitrariness)isProperty-AspectOf signum
linguae (linguistic sign).
x isRepresentativeOf y: States that a person is a representative of a sci-
entific discipline, movement or model.
Example 10. de Saussure (de Saussure)isRepresentativeOf schola Genavensis
(Geneva School).
x isResultOf y: States that an entity falling under a concept is the result
of an entity falling under another concept.
Example 11. vocabulum externum (loan word)isResultOf mutuatio
(borrowing ).
x isSubjectOfDiscipline y: States that a concept that represents some
object (area) is the subject of a concept denoting the scientific discipline or a
theory or model thereof.
Example 12. systema vocabulorum (lexicon) isSubjectOfDiscipline lexicolo-
gia (lexicology ).
x manifests y: States that a concept denotes a grammatical or derivative
category which manifests a concept that denotes a semantic, cognitive, commu-
nicative or functionally determined concept.
Example 13. tempus grammaticum (tense)manifests tempus (time).
x marks y: States that a concept represents a grammatical category which
marks a grammatical relation or function represented by another concept.
Example 14. casus accusativus (accusative)marks objectum directum (direct
Figure 4shows how the modelling of the subordinatingRelation property
results in a taxonomic systematization of Concept instances. This allows for
automatic reasoning over a dataset to yield insights such as ‘nomen’ is super-
ordinate to ‘nomen substantivum’ which is superordinate to ‘nomen commune’
and, thus, ‘nomen’ is also superordinate to ‘nomen commune’. This holds also
for some of the subproperties, e.g. isAKindOf which is a transitive property
(’nomen commune’ isAKindOf ‘nomen substantivum’ and of ‘nomen’). What is
more, the 14 established object properties are all semantically more specific than
a generic “see also” relation but general enough to be broadly applied to interre-
late various (and ideally all) concepts. Especially relations such as isOperatorOf
or marks play a central role in the domain of linguistic terminology. In that
respect, a dataset modelled with OnLiT sets every linguistic term or concept in
a meaningful interrelation to relevant other terms by placing it in a navigable
and coherent context within the linguistic domain a dataset describes. Finally,
relations such as isAKindOf and isPartOf are general across ontologies of any
science and thus serve to integrate linguistic ontologies into an all-encompassing
56 B. Klimek et al.
5 Conclusion and Future Work
The OnLiT data model for representing terminological data of linguistic domains
has been created as the ontological schema basis to transfer the currently rela-
tional database of the LiDo Glossary of Linguistic Terms into an RDF dataset in
the future. Moreover, the OnLiT model constitutes a valuable contribution for
users and creators of linguistic data. Due to the outlined benefits of the under-
lying Linked Data format, evolving terminological data will be interoparable,
semantically and formally explicit as well as easy to reuse and extend. More-
over, OnLiT models linguistic terminology in a meaningful and structured way
that goes beyond a single term definition. I.e. the additional subordinating and
coordinating relations allow to derive coherent and specific insights and knowl-
edge about the conceptualization of linguistic terms in a given language dataset.
Therefore, it can benefit producers of language data in creating their own ter-
minological dataset or in interrelating their data to an existing OnLiT dataset
(e.g. the prospective LiDo RDF dataset). Furthermore, future work includes
an interconnection of the OnLiT model with OntoLex19, which will offer more
possibilities of representing and integrating OnLiT Term and Concept resources
within the domain of lexical language data.
Acknowledgements. This paper’s research activities were partly supported and
funded by grants from the EU’s H2020 Programme ALIGNED (GA 644055) and the
Federal Ministry for Economic Affairs and Energy of Germany (BMWi) for the Smart-
DataWeb Project (GA-01MD15010B).
Bußmann, H., Trauth, G., Kazzazi, K.: Lexikon der Sprachwissenschaft. Taylor & Fran-
cis, London (1996)
Farrar, S.: General Ontology for Linguistic Description (GOLD). The LINGUIST List,
Department of Linguistics, Indiana University (2010)
Farrar, S., Langendoen, D.T.: A linguistic ontology for the semantic web. GLOT Int.
7, 97–100 (2003)
Farrar, S., Lewis, W.D.: The gold community of practice: an infrastructure for linguistic
data on the web. Lang. Resour. Eval. 41, 45–60 (2007)
Goecke, D., L¨ungen, H., Sasaki, F., Witt, A., Farrar, S.: GOLD and discourse: domain-
and community-specific extensions. In: Proceedings of the 2005 E-MELD-Workshop
Good, J., Cysouw, M., Albu, M., Bibiko, H.J.: Can GOLD “cope” with WALS? Retro-
fitting an ontology onto the world atlas of language structures. Max Planck Institute
for Evolutionary Anthropology (2005)
Kamlah, W., Lorenzen, P.: Logische Prop¨adeutik oder Vorschule des vern¨unftigen
Redens. Bibliographisches Institut (B.I.-Hochschultaschenb¨ucher 227/227a),
Mannheim (1967)
OnLiT: An Ontology for Linguistic Terminology 57
Kemps-Snijders, M., Windhouwer, M., Wittenburg, P., Wright, S.E.: Isocat: remod-
elling metadata for language resources. Int. J. Metadata Semant. Ontol. 4, 261–276
Lehmann, C.: Um sistema de documenta¸ao para a ling¨ıstica. Instituto de Letras e
Artes, Pontif´ıcia Universidade Cat´olica do RGS (1976)
Lehmann, C.: Linguistische Terminologie als relationales Netz. In: Knobloch, C.,
Schaeder, B. (eds.) Nomination-fachsprachlich und gemeinsprachlich, pp. 215–267.
Springer, Wiesbaden (1996)
Loos, E.E., Anderson, S., Day, D.H., Jordan, P.C., Wingate, J.D.: Glossary of Linguistic
Terms, vol. 29. SIL International, Dallas (2004)
Schuurman, I., Windhouwer, M., Ohren, O., Zeman, D.: CLARIN concept registry: the
new semantic registry. In: Selected Papers from the CLARIN Annual Conference
2015, pp. 62–79. Link¨oping University Electronic Press (2016)
Wilcock, G.: An OWL ontology for HPSG. In: Proceedings of the 45th Annual Meeting
of the ACL on Interactive Poster and Demonstration Sessions, pp. 169–172. Associ-
ation for Computational Linguistics (2007)
Ontologies are knowledge structures that serve as the foundation for semantic applications. These semantic artifacts organize and classify knowledge to define the semantic relations among the concepts. Some ontologies are constructed for a single natural language, whereas others are built for dual or multi-natural languages. The goal of this study is to design and implement a mobile educational application to help students learn more effectively and to increase the learner’s acquisition of computational linguistics in the Algorithm domain. The methodology is to develop an application with unique features that can display the terms and the concepts in both Arabic and English languages in two different layouts: a list form and a flashcard form. The created app “ArEnAlg” can generate the terms in the algorithm domain. The major objective of the exercise modules is to examine the user’s knowledge of the concepts in the domain of algorithms. The evaluation results tested three factors: the learner’s enjoyment, the acquisition of the computational linguistics, and the application presentation, which showed that behavioral responses support the purpose of the developed application with a max-min range of 99% -81%.
Full-text available
Unter wissenschaftspraktischem Gesichtspunkt ist linguistische Terminologie eine Menge von Schlagwörtern, die z.B. zur Literatursuche nach Sachkriterien benutzt wird. Sie sind durch eine Menge von vordefinierten Relationen miteinander verknüpft, die die Verweisung auf Nachbartermini leisten und formale Aspekte des Begriffssystems wiedergeben. Vermöge ihrer logischen Eigenschaften tragen einige dieser Relationen zur Bildung einer Hierarchie unter den Termini bei. Das entstehende terminologische Netz ist also teilweise systematisch. Es ist, wenn in EDV implementiert, insofern leichter zu pflegen und insbesondere konsistent zu halten als eine Menge von Schlagwörtern, die entweder gar nicht oder bloß durch eine logisch-semantisch leere Verweisrelation verknüpft wären. Dem Benutzer erleichtert es die Orientierung in der Terminologie und unterstützt ihn bei dem Anliegen, einen zusammenhängenden Ausschnitt linguistischer Begrifflichkeit zu bearbeiten.
Full-text available
The Max Planck Institute for Psycholinguistics in Nijmegen, The Netherlands, is creating a state-of-the-art web environment for the ISO TC 37 (terminology and other language and content resources) metadata registry. This Data Category Registry (DCR) is called ISOcat and encompasses data categories for a broad range of language resources. Under the governance of the DCR Board, ISOcat provides an open work space for creating data category specifications, defining Data Category Selections (DCSs) (domain-specific groups of data categories), and standardising selected data categories and DCSs. Designers visualise future interactivity among the DCR, reference registries and ontological knowledge spaces
Full-text available
“database of databases ” consisting of 141 typological databases, covering a wide range of grammatical features, joined into one composite resource through the use of a common metadata scheme. While this metadata scheme ensures interoperability among databases
Conference Paper
The paper presents an OWL ontology for HPSG. The HPSG ontology is integrated with an existing OWL ontology, GOLD, as a community of practice extension. The basic ideas are illustrated by visualizations of type hierarchies for parts of speech.
The GOLD Community of Practice is proposed as a model for linking on-line linguistic data to an ontology. The key components of the model include the linguistic data resources themselves and those focused on the knowledge derived from data. Data resources include the ever-increasing amount of linguistic field data and other descriptive language resources being migrated to the Web. The knowledge resources capture generalizations about the data and are anchored in the General Ontology for Linguistic Description (GOLD). It is argued that such a model is in the spirit of the vision for a Semantic Web and, thus, provides a concrete methodology for rendering highly divergent resources semantically interoperable. The focus of this work, then, is not on annotation at the syntactic level, but rather on how annotated Web resources can be linked to an ontology. Furthermore, a methodology is given for creating specific communities of practice within the overall Web infrastructure for linguistics. Finally, ontology-driven search is discussed as a key application of the proposed model.
  • H Bußmann
  • G Trauth
  • K Kazzazi
Bußmann, H., Trauth, G., Kazzazi, K.: Lexikon der Sprachwissenschaft. Taylor & Francis, London (1996)
General Ontology for Linguistic Description (GOLD) Department of Linguistics (The LINGUIST List)
  • S Farrar
Farrar, S.: General Ontology for Linguistic Description (GOLD). Department of Linguistics (The LINGUIST List), Indiana University (2010)
GOLD and discourse: domain-and community-specific extensions
  • D Goecke
  • H Lüngen
  • F Sasaki
  • A Witt
  • S Farrar
Goecke, D., Lüngen, H., Sasaki, F., Witt, A., Farrar, S.: GOLD and discourse: domainand community-specific extensions. In: Proceedings of the 2005 E-MELD-Workshop (2005)
Logische Propädeutik oder Vorschule des vernünftigen Redens Mannheim: Bibliographisches Institut (B.I.Hochschultaschenbücher 227/227a) (1967) Kemps-Snijders Isocat: remodelling metadata for language resources
  • W Kamlah
  • P Lorenzen
  • M Windhouwer
  • M Wittenburg
  • P Wright
Kamlah, W., Lorenzen, P.: Logische Propädeutik oder Vorschule des vernünftigen Redens. Mannheim: Bibliographisches Institut (B.I.Hochschultaschenbücher 227/227a) (1967) Kemps-Snijders, M., Windhouwer, M., Wittenburg, P., Wright, S.E.: Isocat: remodelling metadata for language resources. In: International Journal of Metadata, Semantics and Ontologies. vol. 4, pp. 261-276 (2009)