Conference PaperPDF Available

The Shortcomings of Language Tags for Linked Data when Modeling Lesser-Known Languages



In recent years, the modeling of data from linguistic resources with Resource Description Framework (RDF), following the Linked Data paradigm and using the OntoLex-Lemon vocabulary, has become a prevalent method to create datasets for a multilingual web of data. An important aspect of data modeling is the use of language tags to mark lexicons, lexemes, word senses, etc. of a linguistic dataset. However, attempts to model data from lesser-known languages show significant shortcomings with the authoritative list of language codes by ISO 639: for many lesser-known languages spoken by minorities and also for historical stages of languages, language codes, the basis of language tags, are simply not available. This paper discusses these shortcomings based on the examples of three such languages, i.e., two varieties of click languages of Southern Africa together with Old French, and suggests solutions for the issues identified.
The Shortcomings of Language Tags for Linked
Data When Modeling Lesser-Known Languages
Frances Gillis-Webber
Library and Information Studies Centre, University of Cape Town, South Africa
Sabine Tittel
Heidelberg Academy of Sciences and Humanities, Germany
In recent years, the modeling of data from linguistic resources with Resource Description Framework
(RDF), following the Linked Data paradigm and using the OntoLex-Lemon vocabulary, has become
a prevalent method to create datasets for a multilingual web of data. An important aspect of data
modeling is the use of language tags to mark lexicons, lexemes, word senses, etc. of a linguistic
dataset. However, attempts to model data from lesser-known languages show significant shortcomings
with the authoritative list of language codes by ISO 639: for many lesser-known languages spoken
by minorities and also for historical stages of languages, language codes, the basis of language tags,
are simply not available. This paper discusses these shortcomings based on the examples of three
such languages, i.e., two varieties of click languages of Southern Africa together with Old French,
and suggests solutions for the issues identified.
2012 ACM Subject Classification
Computing methodologies
Language resources; Information
Dictionaries; Information systems
Semantic web description languages; Information
Graph-based database models; Information systems
Resource Description Framework
(RDF); Software and its engineering
Interoperability; Information systems
Multilingual and
cross-lingual retrieval; Computing methodologies
Information extraction; Computing methodolo-
gies Artificial intelligence
Keywords and phrases
language codes, language tags, Resource Description Framework, Linked
Data, Linguistic Linked Data, Khoisan languages, click languages, N|uu, k’Au, Old French
Digital Object Identifier 10.4230/OASIcs.LDK.2019.4
We would like to thank the reviewers for helpful comments and insightful
1 Introduction
The publication of language data on the Web as Resource Description Framework (RDF), and
according to Tim Berners-Lee’s Linked Data principles
, has contributed to the emergence of
a multilingual web of data. Publishing language resources as Linked Data allows for language
resources to be exploited with the benefits of structural interoperability (same format and
query language leading to cross-resource access), conceptual interoperability (shared standard
vocabularies), accessibility (via standard Web protocols), and resource integration (via linked
resources) [6].
1 [10-01-2019].
©Frances Gillis-Webber and Sabine Tittel;
licensed under Creative Commons License CC-BY
2nd Conference on Language, Data and Knowledge (LDK 2019).
Editors: Maria Eskevich, Gerard de Melo, Christian Fäth, John P. McCrae, Paul Buitelaar, Christian Chiarcos,
Bettina Klimek, and Milan Dojchinovski; Article No. 4; pp. 4:1–4:15
OpenAccess Series in Informatics
Schloss Dagstuhl – Leibniz-Zentrum für Informatik, Dagstuhl Publishing, Germany
4:2 Shortcomings of Language Tags for Lesser-Known Languages
After a brief introduction to RDF and Linked Data, particularly in the context of
linguistic resources, as well as language codes and language tags (Section 1), we present the
challenge addressed in this paper: finding solutions for the shortcomings of language tags
when identifying near-extinct and historical languages (Section 2), and we do so by modeling
data from three languages, e.g., two click varieties from the language family previously
referred to as ‘Khoisan’, and Old French (Section 3). The paper concludes with a discussion
of the findings (Section 4) and directions for future work (Section 5).
1.1 RDF and (Linguistic) Linked Data
RDF is the standard data model for resources of the Semantic Web [
]. It expresses data as
subject-predicate-object triples to facilitate data interchange on the web. Each subject and
object is a node; the predicate forms a relation (edge) between two nodes. The subject can be
a URI (Uniform Resource Identifier) or a blank node, the predicate can only be a URI, and
the object can be a URI, blank node or a literal (described as a string), see [9, 3].
Linked Data (LD) can be defined as the «set of best practices for publishing and connecting
structured data on the Web», and it builds on the RDF data model using HTTP (Hypertext
Transfer Protocol) URIs [
, 4-12]. The LD principles have been adapted in many fields,
including linguistics, where it has led to the creation of numerous datasets published as
Linguistic Linked Open Data (LLOD)
: lexicons, annotated corpora, dictionaries, etcetera
]). The model that has become the de facto standard for describing linguistic resources
is the OntoLex-Lemon vocabulary
, 587]. The focus within this field lies on well-resourced
languages and, in particular, on their modern stages, with a small number of examples
of linguistic resources documenting low-resourced languages (e.g., [
]) and also
historical language stages (e.g., [32, 7, 22, 31, 2]).
1.2 Language codes language tags
To use unique codes for the identification of languages is necessary for any environment that
follows BCP 47 [
]. Examples include language identification in RDF and XML documents
(the latter using the
attribute), and institutions such as language repositories,
e.g., the Open Language Archives Community (OLAC) and the World Atlas of Language
Structures (WALS).
A unique language code is able to disambiguate the case when one
language name refers to several languages, and one language has several names.
A language code «represents one or more language names, all of which designate the same
specific language» [
]. The International Organization for Standardization (ISO) provides a
standard for language codes: ISO 639 with Parts 1–3. In principle, the language codes in
each part «are open lists that can be extended and refined», and a Registration Authority
nominated by ISO maintains each part [
]. ISO 639-1 provides a two-letter code and it is
a subset of ISO 639-2, which provides a three-letter code allowing for more languages to
be represented. Both ISO 639-1 and ISO 639-2 represent major languages that are most
frequently expressed in the world’s literature ([
]). The individual languages in ISO
639-2 are in turn a subset of those in ISO 639-3 that aims «to give as complete a listing of
languages as possible» [
]. The types of languages covered include living, extinct, ancient,
historic and constructed languages; their scope can either be an individual language or a
macrolanguage, and the modality is spoken, written or signed ([12, 18, 19, 20]).
2 [26-12-2019].
3 [31-12-2018].
4; [15-03-2019].
F. Gillis-Webber and S. Tittel 4:3
For individual languages, only varieties which are considered to be distinct languages are
represented in ISO 639-3, with any dialects encompassed within the language code of that
language. The language code «represents the complete range of all the spoken or written
varieties of that language, including any standardized form» [
]. A macrolanguage code
represents a cluster of language varieties. Macrolanguages differ from language collections in
that for the former, the languages must be deemed very closely related, and for the latter,
there can be a loose relation, but there should be some connecting feature, be it historical,
geographical, or a linguistic association [
, 33]; language collections are only represented in
ISO 639-2, and macrolanguages are only represented in ISO 639-3 [19].
A language tag is similar in concept to a language code, except the latter can be used
in any discipline, and the former is intended for the internet community. The scope of a
language tag is defined by IETF’s BCP 47. BCP 47 is a document which specifies Best
Current Practice for tags for identifying languages, and the language in question is able to
be refined further from the ISO 639 language code ([
]; [
, 1-4]; [
]). Language tags are
of the form: language-extlang-script-region-variant-extension-privateuse, comprised of one or
more sub-tags, each separated by a hyphen; language is the shortest language code from ISO
639, and the remaining sub-tags are distinguished from each other «by length, position in
the tag, and content» ([21]; [28, 4]).
1.3 Language codes for linguistic resources
The OntoLex-Lemon specification requires each linguistic resource, be it a lexicon, a lexical
entry, or a lexical concept, to be identified using a URI to the relevant ISO 639 code, with
RDF requiring each string literal in an object to be ‘language-tagged’.5
A language code (or tag) is thus used in the following scenarios:
1. to identify a lexicon:
when a triple with the predicate
is declared: this is to the URI of an
ISO 639 language code [8];
2. to identify a lexical entry:
same as (1);
3. for the language tagging of string literals:
this is a language tag, which, in the absence of additional sub-tags, is an ISO 639
language code [9].
A lexical entry in RDF, described using OntoLex-Lemon and serialized in Turtle
, can be
modeled as follows:
1@PREFIX ontolex:< ht t p : // w w w . w3 . o r g / ns / l e m on / o nt o l ex # > .
2@PREFIX lexinfo:< ht tp : // w w w . le x in f o . n et / o n to l o gy / 2. 0 / l ex i nf o # > .
3@PREFIX dct:< ht t p : // p u rl . or g / d c / t er m s / > .
4@PREFIX rdfs:< ht t p : // w w w . w3 . o r g / 2 00 1 / 0 2/ r d f - s c he m a # > .
6: en t ry / en - n- b i le a ontolex:LexicalEntry , ontolex: W o rd ;
7lexinfo:partOfSpeech lexinfo: N ou n ;
8dct:language < ht tp : // i d . lo c . g ov / v o ca b u la r y / i so 63 9 - 2/ e n > ,
9< ht t p : // l e xv o . o rg / i d / is o 63 9 -1 / e n > ;
10 rdfs: l ab e l " bi l e "@ e n ;
5 this-document [10-01-2019].
Beyond RDF, OntoLex-Lemon, and DublinCore (dct) vocabulary, we use classes and properties of
LexInfo, RDFS, SKOS, and DBpedia, see the respective URLs within the code examples.
Terse RDF Triple Language, an easy to read serialization of RDF statements,
turtle/ [11-01-2019].
LDK 2019
4:4 Shortcomings of Language Tags for Lesser-Known Languages
11 ontolex: c a no n ic a l Fo r m : en t ry / en - n- b i le # l em m a ;
12 ontolex: s en s e : en t ry / en - n- b i le #sense1 ;
13 ontolex: e v ok es : c on ce p t / 00 0 00 0 00 1 .
Point 2 is demonstrated in Line 8-9: the applicable language codes for the lexical entry,
from ISO 639-2 and ISO 639-1 respectively, are indicated as ‘English’.
Point 3 is demonstrated in Line 10: the language of the literal “bile” is specified with the
ISO 639-1 code for English.
2 The shortcomings of language tags
The ISO 639 standard list includes more than 6,900 language codes
but it neither covers all
the world’s languages nor all historical language stages of the languages. This is problematic
when modeling under-resourced or extinct languages for which a language code does not
exist. To the best of our knowledge, this problem has not been properly addressed in the
literature. A recent email thread in the W3C Semantic Web forum
expressed the opinion to
do away with language tags altogether, but there was not shared consensus on this point.
Chiarcos and Sukhareva [
] show the conversion of legacy data from dictionaries of the
historical language stages of Germanic languages (Old Saxon, Old High German, Old Norse,
etc.) and find the following compensation for the lack of language codes within ISO 639:
they preserve the original language abbreviations of the dictionary resource and extract «all
language identifiers, and by a hand-crafted mapping from the original abbreviations», ISO
639-3 codes are assigned where possible [
, 44b]. The language URIs are represented using
lexvo [
], but «[u]nfortunately, many abbreviations could not be resolved against lexvo, in
particular, this included hypothetical forms for reconstructed historical language stages, e.g.,
Proto-Germanic.» They conclude that the extension of existing terminologies with respect to
historical language stages is a great desideratum [
, 44b]. Their approach results in code such
lemon:language "ae."@deu
, with ‘ae. being the German abbreviation for Altenglisch in
the dictionary resource [
], see [
, 44b], and ‘deu’ being the ISO 639-3 language code for
Standard German.10
The same approach has been taken by Declerck et al. for the transformation of the data
from the Wörterbuch der bairischen Mundarten in Österreich (WBÖ)
into LD [
]: in
the code sample given at [
, 347], the language tag for the Bavarian language is modeled
as a literal:
, which raises the question why it is not given in the form
, ‘bar’ being the ISO 639 code for Bavarian.
One might speculate that this could
serve as a means to distinguish the language documented in the WBÖ (Bavarian varieties
spoken in Austria) from Bavarian spoken in Bavaria; however, the problem is not addressed
in the paper.
Amongst the findings of Tittel and Chiarcos [
] is the fact that due to the lack of
appropriate language codes, the problem of modeling the different dialectal forms of lexemes
in linguistic resources of Old French is still unsolved: for the conversion of the data of
8According to the table in [10-01-2018].
Language-tagged strings Re: Towards easier RDF: a proposal [Electronic mailing list, 23-26 Novem-
ber] 2018,
10 [11-01-2019] (639-1: ‘de’; 639-2/B: ‘ger’).
11 [11-01-2019].
12 [11-01-2019].
F. Gillis-Webber and S. Tittel 4:5
the Dictionnaire étymologique de l’ancien français (DEAF, [
]) following the Linked Data
paradigm, the researchers established that all graphical variants of a given Old French lexeme
could only be identified by ISO 639-3 code ‘fro’
for overall Old French. This meant that
information originally included in the linguistic resource such as ‘Anglo-Norman’ or ‘medieval
Lorraine’ scripta
– information that is very valuable for the research of Old French dialects
– would be excluded from the language description when converted to Linguistic Linked
Data. To solve this problem, [
, 65] propose to define the code ‘fro’ in ISO 639-3 as a
macrolanguage and to register the Old French dialects as varieties associated to ‘fro’. (There
had been an attempt to include varieties of historic languages within ISO 639-6, but this
Part was withdrawn in 2014.15)
Bellandi et al. [2] discuss the modeling of linguistic data from Old Occitan (a Romance
language spoken during the Middle Ages in what is today southern France) and other
languages using OntoLex-Lemon. To code their Old Occitan lexemes, they use the tag
lemon:writtenRep "canabo"@aoc
, 4]. One rightly assumes that this is the ISO
639-3 code ‘aoc’, however, ‘aoc’ represents the Pemon language of the Cariban language
family, a language in Venezuela.
The correct ISO 639 code for the language is ‘pro’ (=
Old Provençal, the former term for the language)
, and presumably ‘aoc’ simply is an
abbreviation for French ancien occitan. Their handling of the use of codes is illegal: the
definition of a language tag using the ‘@’ sign and a language code must be BCP 47 compliant
to be valid.
] do not address this issue, nor do they address the issue of creating their
own language codes.
We conclude that new language codes need to be created, in a way that adheres to current
standards and best practices of language identification. The objective of this paper is to
contribute to the discussion of this problem. On the basis of three example languages, we
will propose solutions to meet the requirements of the languages discussed.
The following languages serve as our examples:
uu and
’Au: two dialects from N
ng, a critically endangered non-Bantu click language
in Southern Africa, that are both near-extinct [30, 7].
2. Old French: the ancestor of modern French, spoken during the Middle Ages.
In our sample code, we will focus on language-tagged string literals. It is clear, however,
that the described problems and proposed solutions also apply to language URIs for lexicons
and lexical entries.
3 Finding solutions for N|uu and k’Au, and Old French
We focus on varieties of N
ng and Old French to underline the fact that they are good
examples of the need to preserve the languages and their historical stages as a key to
understanding our cultural heritage: language is the storehouse of our culture, both past
and present. It captures all aspects of life. It is subject to change and, thus, mirrors the
development of our culture, of our state of mind, and of our social interaction through time.
13 [07-01-2019].
Scripta is the term for the written form of a spoken dialect. Anglo-Norman is one of the varieties of Old
French; it was spoken in England during the Anglo-Norman period.
15 [07-01-2019].
16, [11-01-2019].
17 [10-01-2019].
18 Graph-Literal
[11-01-2019]. – Note also that ‘@arab’ is used to represent Arabic,
although the ISO 639 code is ‘ara’, [11-01-2019].
LDK 2019
4:6 Shortcomings of Language Tags for Lesser-Known Languages
As little connected as N
’Au, and Old French might ostensibly seem, they serve well
to illustrate the problem: ISO 639 codes do not exist for N
’Au, and the varieties of
Old French. The atypicality of these languages highlights the relevance of the problem on a
broader scale: more under-resourced, extinct or historical languages that are (currently) not
included in the ISO 639 language code list will be published as LLOD.
3.1 N|uu and k’Au
ng is the name of a dialect cluster of the !Ui-Tuu language family (formerly referred to as
Southern Khoisan), spoken over a geographically large area in the southern Kalahari Desert;
uu is the Western variety of N
ng, and
’Au, the Eastern variety ([
, 11-17]; [
]; [
, 27]).
Both dialects are near-extinct with two speakers for
’Au and three speakers for N
uu as of
2013 (with the most fluent speaker of N
uu acting as a language teacher to young people); all
ng speakers use Afrikaans as their main language [
, 15-16]. Since the late 19th Century,
linguists have collected data of Khoisan
languages: this data is sparse, heterogeneous and
difficult to access with misclassified languages, inappropriate language names and insufficient
metadata as examples of the challenges faced, in addition to the identity of diverse corpora in
archival material hard to assess, both in relation to each other and to modern languages ([
5-8]; [
, 2]). To document the many Khoisan languages is a challenge and a desideratum at
the same time: encoding data following the Linked Data paradigm will convert the data into
a valuable resource, possibly giving way to linguistic reconstructions using computational
methods, where standard linguistic methodologies have been unable to yield meaningful
results [
, 1]. Making accessible and preserving this data will contribute significantly to
the exploration of the cultural heritage of mankind, with the collective group of Khoisan
speakers being one of the few remaining hunter-gatherer cultures worldwide and the oldest
existing human group today, according to genetic studies [29, 379].
3.1.1 Existing language codes
In order to convert the linguistic data of N
uu and
’Au resources, we need an appropriate
means to denote the languages in an unambiguous way, i.e., language codes to label the
modeled elements of the linguistic resources in RDF. A language code for N
ng exists, i.e.,
ISO 639-3 ‘ngh’; this code is shared by both sub-languages N
uu and
according to the archival Khoisan ‘doculects’ discussed by [
, 16], the differences between
the two language varieties are significant and, thus, explicit language codes for both
and N|uu are required.
Within MultiTree, a library of language relationships hosted by The Linguist List, the
codes for N
uu and
’Au are ‘ngh-nuu’ and ‘ngh-aun’ respectively.
Both are documented
for ‘Private Use’, however their syntax does not meet the requirement defined by IETF’s
BCP 47, where the private use portion of the tag must be prepended with ‘
’ ([
]; [
4]). Furthermore, both the latter portions of MultiTree’s codes, namely ‘nuu’ and ‘aun’, are
pre-existing language codes, i.e., the former for the language Ngbundu (a language of the
Congo area), and the latter for Molmo One (Papua New Guinea).
Despite the fact that
The modern Khoisan languages are classified into three families and two isolates: the families Kx’a,
!Ui-Tuu and Khoeid, and the isolates Hadza and Sandawe [5, 2].
20 [29-12-2018].
21,.../codes/ngh- aun [20-06-2018].
22 registry/language-subtag-registry
F. Gillis-Webber and S. Tittel 4:7
the use of privateuse sub-tags is by definition by private agreement only (cf. Point
of BCP 47, [
, 18]), it is clear that the use of MultiTree’s language tags ‘ngh-nuu’ and
‘ngh-aun’ may lead to inadvertent misinterpretation when included in a language tag.
For this reason, we consider the use of Glottolog, a comprehensive catalogue of the world’s
lesser-known languages maintained by the Max Planck Institute for the Science of Human
History. Their catalogue «assigns a unique and stable identifier (the Glottocode) to (in
principle) all languoids, i.e. all families, languages, and dialects», [
]. Glottolog registers
the two languages N
u and
’Au (as sub languages of N
with the codes ‘nuuu1242’ and
‘auni1243’, respectively. However, as BCP 47 only allows for ISO 639 language codes in its
language sub-tag, Glottolog is not recognized as a standard.
3.1.2 The use of privateuse sub-tag
In light of unambiguous language codes being available for the two Khoisan varieties, we
propose to combine the ISO 639-3 code for the parent language N
ng, i.e., ‘ngh’, with the
privateuse sub-tag ‘x-’ and the respective Glottocodes stated above.
The language tags for N|uu and k’Au can then be defined accordingly:
N|uu: ngh-x-nuuu1242
k’Au: ngh-x-auni1243
A lexical concept, which can be linked to one or more senses in lexical entries from
different languages, can be modeled as follows:
1@ P RE F IX s ko s : < ht t p : // ww w . w 3 . or g / 2 0 04 / 0 2 / s ko s / c o re # > .
2@PREFIX dbr:< ht tp : // d b p ed i a . or g / r e so u rc e / > .
5: c on c ep t / 0 00 0 0 00 0 1 a sk os : C o nc e pt , ontolex: LexicalConcept ;
6sk o s : ex a mp l e " T he b el l y is f at " @ e n ;
7sk o s : ex a mp l e "k’ â he !q h û i a ." @ n gh - x - n uu u 1 24 2 ;
8ontolex: l e xi c a li z ed S en s e : en - n - be ll y #sense1 ;
9ontolex: l e xi c al i ze d Se n se : n g h_ x_ n uu u 12 4 2 - n - xa _b e ll y #sense2 ;
10 ontolex:isConceptOf dbr:Abdomen .
Lines 6-7 show language-tagged strings, and line 7 the compiled language tag for N|uu.
3.2 Old French
Old French is the French spoken in the Middle Ages, and it can be more precisely defined
as the umbrella term for the different Old French dialects
spoken in what is now France,
parts of Belgium, England, Italy and the Holy Land. Its written resources date from 842
AD until c. 1350 AD (the border with Middle French) and its remarkable written tradition
serves to document its role as the most important vernacular of this time in Europe.
23 [24-06-2018].
The DEAF registers 30 varieties of Old French, Franco-Italian (a written, artificial language in the
Middle Ages), and Judeo-French (sociolect), see Table 3, Appendix.
Approx. 3,000 primary text sources transmitted within more than 10,000 manuscripts are registered by
the Complément bibliographique of the DEAF,
LDK 2019
4:8 Shortcomings of Language Tags for Lesser-Known Languages
3.2.1 Existing language codes
BCP 47’s language tag offers a variant sub-tag that can be «used to indicate additional,
well-recognized variations that define a language or its dialects that are not covered by other
available subtags», where one or more variants can be used to form a language tag. Each
of these variant sub-tags must be registered with IANA before use [
, 15]. Middle French
is registered (ISO 639-3 code ‘frm’)
but no variants have been registered for Old French.
IANA has registered Anglo-Norman (ISO 639-3 code ‘xno’), but not as a sub-category of
Old French, although it should be considered as such; the same applies to Zarphatic (‘zrp’:
Judeo-French, spoken in the Middle Ages).
MultiTree lists Old French (‘fro’) and also the following child languages: Picard (ISO 639-3
code ‘pcd’), Walloon (‘wln’), and Zarphatic (‘zrp’); Anglo-Norman (‘xno’) is not registered
as a child language.
Although Walloon is registered as a child language of Old French, it is
described as a living language; the same applies to Picard. Middle French is also registered
as a child language of Old French, thus, following this logic, so should modern French. The
hierarchization of Judeo-French (variety of Old French: sociolect) on the same level as Middle
French (successor of Old French) and Picard / Walloon (modern dialects of the Picardy and
Wallonia, respectively) conflates synchronic, diachronic, and geographical aspects.
Glottolog has assigned the identifier ‘oldf1239’ to Old French
but Glottolog does not
register dialects of the medieval time period.
In addition to this flaw, Glottolog does not
seem appropriate for the needs of linguists modeling data from the Romance languages,
particularly with regard to old language stages. A closer look at Glottolog reveals major
shortcomings in both the registration and the hierarchization of the Romance languages. E.g.,
Glottolog conflates diachronic and dialectal criteria within its hierarchies in several ways:
Old French is registered (as a sub-entity of ‘Oïl’
) at the same level as modern ‘Central Oïl’,
Francoprovençalic (Romance language spoken in Eastern France), and Walloon. Following
the hierarchy into the branches and sub-branches of ‘Central Oïl’ we find
Global French
a number of modern French dialects, but, also, Middle French and
We deem necessary a thorough revision of the hierarchies, (re-)assembling
both the dialects and regional varieties of modern French, and the historic stages of French.
3.2.2 Preliminary findings
The evaluation of language tags and language hierarchies in ISO 639, BCP 47, IANA,
MultiTree, and Glottolog shows that the assignation of language codes to Old French
dialects is not straightforward. At least for Anglo-Norman and Zarphatic, which we consider
sub-categories of Old French, ISO 639-3 provides codes, i.e., ‘xno’ and ‘zrp’ respectively.
These codes can be used for modeling lexemes and their graphical variants characterized as
Anglo-Norman or Zarphatic. The following example for the Anglo-Norman noun firbote
illustrates this:
The sub-tag ‘frm-1606nict’, templates/1606nict.
[08-01-2019], does not depict a regional variety but the language documented by Jean Nicot in his
Thresor de la langue françoyse, tant ancienne que moderne, Paris, from 1606.
27;.../pcd;.../wln;.../zrp;.../xno [07-01-2019].
28 [07-01-2019].
Old French is not available in the language collection of Ethnologue, as «ancient, classical, and long-
extinct languages are not listed»,[29-12-2018].
The term for the Romance varieties using an adaptation of the Vulgar Latin term hoc ille “this (is) it”
as ‘Yes’.
31 More modern French dialects are found scattered in other sub-branches.
Juridical term (in England) designating the right to take firewood from the land of a landlord, DEAF F
492,29, https://deaf-server.adw.uni- [08-01-2019].
F. Gillis-Webber and S. Tittel 4:9
1<firbote> aontolex:LexicalEntry , ontolex: W or d ;
2lexinfo:PartOfSpeech lexinfo: N ou n ;
3ontolex: c a no n ic al F or m <f i r bo t e # f or m > .
5< f ir b ot e # f o rm > aontolex : Fo r m ;
6ontolex: w r it te n Re p "firbote"@ xn o .
3.2.3 The use of privateuse sub-tag
For the other Old French dialects and language varieties (see Table 3, Appendix), as language
codes are not available, we again have to consider the use of BCP 47’s privateuse sub-tag. E.g.,
a tag for the Old French variety spoken in Lorraine, a region in north-eastern France, could
be defined as
. A simple example of an Old French word form characteristic
of the Lorraine scripta is feyvre, a graphical variant of Old French fevre m.
This can be
modeled as follows:
1< fe v r e > aontolex:LexicalEntry , ontolex : Wo r d ;
2ontolex: c a no n ic al F or m < f ev r e # f o rm _ 1 > ;
3ontolex: o t he rF o rm < fe v r e # f or m _ 2 > .
5# Ol d Fr e nc h s ta n da r d fo r m (l e mm a )
6< fe v r e # f or m _ 1 > aontolex: Fo r m ;
7ontolex: w r it te n Re p " fe v re " @ f ro .
9# gra p hi c al var i an t
10 < fe v r e # f or m _ 2 > aontolex: Fo r m ;
11 ontolex: w r it te n Re p "feyvre"@ fr o - x - l o rr a i ne .
3.2.4 Adding geographic information
The language tag can be further enriched by including geographic information, in line with
established standards. There are several options available to us: (1) we could refer to
the administrative region of France, (2) to the French département, or (3) use geographic
coordinates. Both the administrative region and the département can be identified using the
codes of the ISO 3166 standard for the administrative subdivisions of France.34 Administrative region and département
The area ‘Lorraine’ is part of the region Grand-Est (covering Alsace, Champagne, Ardenne,
and Lorraine), thus the language tag can be defined as
the administrative region covers an area considerably larger than the geographic area of
Lorraine, and thus does not map the area in question in a satisfying way. Another option
would be to enrich the language tag by referring to the département, which would allow us
to map the area more precisely.
Regarding options (1) and (2), the following concerns are raised:
The smith, DEAF F 342,21,
34 [07-01-2019].
35 Ibid.
LDK 2019
4:10 Shortcomings of Language Tags for Lesser-Known Languages
The administration of regions and départements is subject to change. As a consequence,
the ISO 3166 codes are unstable, as evidenced by sub-divisions being allocated to new
metropolitan regions in France as recently as 2016.36
The area in which an Old French dialect was spoken can embrace several modern
regions, e.g., ‘Nord-Est’ and ‘Sud-Ouest’ (see Table 3, Appendix), or départements:
e.g., contemporary Lorraine consists of not one but four départements, i.e., Meurthe-et-
Moselle (ISO 3166-2:FR-54), Meuse (ISO 3166-2:FR-55), Moselle (ISO 3166-2:FR-57),
and Vosges (ISO 3166-2:FR-88); the historical region also comprises the contemporary
département Haute-Marne (ISO 3166-2:FR-52). As a result, either more than one
region may need to be included in the sub-tag, indicating (imprecisely) the geographical
boundary in which the dialect was spoken, or the RDF triples must be manifolded: when
modeling a lexeme or a graphical variant of a lexeme characterized as Lorraine, e.g.,
within the data of the DEAF dictionary, the inclusion of the codes for the départements
into the language tag requires duplicating the RDF triples, thus creating somewhat
unwieldy data.
The boundaries of the regions are modern-day boundaries which may not necessarily
align to the boundaries of a previous time. This leads to a dissatisfying mapping of
said area. Geographic coordinates
As a third option, we consider the inclusion of geographic coordinates in the language tag.
To do this, we map the (approximate) geographic distribution of ‘Lorraine’ to coordinates,
assuming that the last coordinate is the same as the first coordinate, and the coordinates are
ordered in a counterclockwise direction, thus creating a polygon shape [
]. Each coordinate
can be compressed using Geohash, a system for encoding geographic coordinates into a base32
string, which would also format each latitude and longitude value in a syntax acceptable
for BCP 47.
As precision down to the nearest meter is not necessary, the Geohash length
could be limited to five characters,
rendering the coordinate in an approximate area that is
4.89 x 4.89 kilometers.39
As using the geographic coordinates to map the modern-day distribution of ‘Lorraine’
would lead to the same dissatisfying result (cf., we draw on a map of the Französisches
Etymologisches Wörterbuch – FEW [34] that includes historical information, see Fig. 1.40
To derive the geographic coordinates for the old dialect of ‘Lorraine’, we take this map as
a substratum and the result is the following:
(4.91473,49.62686), (4.6696405,48.0428789), (5.59192,47.6435), (6.858446002006532,
47.883257283545234), (7.2386756,48.4086571), (5.81263,49.72584), (4.91473,49.62686)41
36 [29-12-2018].
Geohash 2018,
[31-12-2018]; Geo-shape datatype
38 Or less, depending on the extent of the geographical distribution of the dialect being mapped.
39 [31-12-2018].
In the possession of the editorial office of the DEAF is a 40-year-old, battered copy of the map of France
that is included in the Beiheft of the FEW. This copy contains the boundaries of the areas where the
Old French dialects were spoken, sketched in by hand (in yellow) by Frankwalt Möhren, co-founder of
the DEAF (and, also, valuable notes and comments, e.g., the indication ‘Orval: Bier!’: the Abbey of
Orval in Villers-devant-Orval is the home of the famous top-fermented beer ‘Orval’).
41 Ordering is latitude then longitude.
F. Gillis-Webber and S. Tittel 4:11
Figure 1
‘Old Lorraine’ area: Extract of the map of the FEW (left), mapped using geographic
coordinates (right).
Each longitude and latitude coordinate can be converted to a Geohash, to a precision of
five characters: t0g7c, t0f4t, t0czu, t14p1, t163j, t1535, t0g7c.
As the last coordinate is the same as the first coordinate, the last one can be excluded,
and as only alphanumeric characters and hyphens are allowed by BCP 47, every Geohash,
with the exception of the first one, is prepended with ‘
’ to serve as an internal delimiter;
the language code for the language, dialect and region can thus be presented as follows:
The use of a historical map as a source of information to enrich a language tag with
geographic coordinates, as demonstrated for medieval Lorraine, seems very promising to us
regarding our aim: the unambiguous and historically-correct tagging of languages.
A further possibility is to include the period of time within the language tag, e.g.,
where 850AD--1350AD depicts the time range.42
BCP 47 specifies the maximum length of a sub-tag to be of eight characters (+two
for ‘
’, see [
, 6]). However, numerous examples of the privateuse sub-tags exceed this
maximum length [
, 56,81]. Thus, we conclude that there is not an upper limit to the length
of the privateuse sub-tag, except that pertaining to buffer overflow [28, 63,71-72].
4 Discussion
The examples, N
uu and
’Au, and Old French, demonstrate that there is not a single,
encompassing solution that can be applied to all languages. For each of the three languages,
a custom approach, in conjunction with the privateuse sub-tag from BCP 47’s language tag,
has had to be adopted. However, with each example, a tentative pattern for the privateuse
sub-tag has emerged: each part within the privateuse sub-tag can be assigned to a category,
as listed in Table 1, and the privateuse sub-tag can consist of one or more parts.
In the case of Old French, this seems dispensable since the code ‘fro’ contains this information, however it
could be valuable when identifying a language where the geographical distribution changes significantly.
LDK 2019
4:12 Shortcomings of Language Tags for Lesser-Known Languages
Table 1 The categorization of parts in a privateuse sub-tag.
Part Description
language A language, dialect or pidgin not in ISO 639
otherlect An ethnolect, sociolect, or idiolect
If not modern-day; not equivalent to the time period
specified by the language code
region A geographic, politic or administrative region
Using the categories identified in Table 1, we thus propose the following pattern for the
privateuse sub-tag of a language tag, with each part separated by a ‘-’:
Within BCP 47, the format of the language tag has been designed such that each sub-tag
can be identified on the basis of its length, position in the tag, and its content, and each
sub-tag is typically a code from an ISO standard or registry [
, 8]. However, this requirement
can be limiting and inflexible. In order to identify each part in the privateuse sub-tag pattern,
we propose prepending each part with a key consisting of 2 digits, from 0 - 9, with the first
digit, Key 1, indicating the category, and the second digit, Key 2, indicating the content
in relation to Key 1, as shown in Table 2. This way, each part can be of variable length,
thus allowing for greater flexibility. For example, a part that is categorized as language can
be prepended with ‘10’, where ‘1’ indicates that it is language and ‘0’ indicates that the
language is user-defined information. The tags can, thus, be rewritten as follows:
N|uu dialect: ngh-x-01nuuu1242
k’Au dialect: ngh-x-01auni1243
Old French, Lorraine dialect:
Table 2 The key for each part of the privateuse sub-tag.
Part Key 1 Key 2
language 0 0 = User-defined
1 = Glottocode
otherlect 1 0 = User-defined
1 = Glottocode
timeperiod 2 0 = one year only, BC
1 = one year only, AD
2 = start:BC - end:BC
3 = start:BC - end:AD
4 = start:AD - end:AD
region 3 0 = Geohashed latitude and longitude coordinates – polygon
1 = Geohashed latitude and longitude coordinates – point only
2 = URI to GeoJSON-LD
3 = Code from ISO 3166
The interpretation of a language tag which contains multiple sub-tags can be obscure
and requires human inspection. By (1) categorizing the privateuse sub-tag into parts, then
(2) defining a key for each part, and (3) defining rules for each key, it not only allows for
more accurate interpretation, by both human and machine, but it can also lead to increased
shared agreement for a compiled language tag.
F. Gillis-Webber and S. Tittel 4:13
5 Conclusion and Future Work
In this paper, we have discussed the shortcomings of language tags in the context of modeling
data from lesser-known languages as LD. For two under-resourced language varieties and one
historical language stage we have proposed solutions using the privateuse sub-tag, with the
addition of geographic information. This can improve a language tag so that it reflects the
diachronic, synchronic and dialectal aspects of the language in question.
The proposed rule-based pattern for the privateuse sub-tag is not intended to be used in
place of other sub-tags in the language tag, nor is it intended to replace the work of existing
standards and bodies. The W3C Internationalization (i18n) Interest Group
serves to
connect a large group of people on the topic of internationalization on the Web. The authors
intend to contribute to the discussions of the group, submitting the proposals outlined in this
paper for further feedback. Also, the authors and C. Maria Keet propose MoLA, a
del for
nnotation (
) [
]. MoLA has been
developed to provide a vocabulary for language annotation in RDF, which enables custom
language tags to be defined, and for said language tags to be associated with both a time
period and region.
Defining a pattern for the privateuse sub-tag can lead to discussions which can improve
the next iteration of BCP 47, as well as to increased interoperability within the context of
LLOD so as to render language identification more accurate. This in turn can lead to shared
agreement between lexical resources and to re-use, an important notion in a multilingual
Semantic Web.
K. Baldinger. Dictionnaire étymologique de l’ancien français – DEAF. Presses de L’Université
Laval / Niemeyer / De Gruyter, Québec/Tübingen/Berlin, since 1971. [Continued by Frankwalt
Möhren, and Thomas Städtler; DEAFél:https://deaf-].
A. Bellandi, E. Giovannetti, and A. Weingart. Multilingual and Multiword Phenomena in a
lemon Old Occitan Medico-Botanical Lexicon. Information, 9 (3), 52, 2018.
3T. Berners-Lee. Linked Data. World Wide Web Consortium, 2006.
Ch. Bizer, T. Heath, and T. Berners-Lee. Linked Data – The Story So Far. International
Journal on Semantic Web and Information Systems, 5:1–22, 2009.
M. Brenzinger. The twelve modern Khoisan languages. In A. Witzlack-Makarevich and
M. Ernszt, editors, Khoisan Languages and Linguistics: Proceedings of the 3rd International
Symposium July 6-10, 2008, Riezlern/ Kleinwalsertal, pages 1–32. Köppe Verlag, 2008.
Ch. Chiarcos, J. McCrae, Ph. Cimiano, and Ch. Fellbaum. Towards Open Data for Linguistics:
Lexical Linked Data. In A. Oltramari et al., editor, New Trends of Research in Ontologies and
Lexical Resources: Ideas, Projects, Systems, pages 7–25. Springer, Berlin, Heidelberg, 2013.
Ch. Chiarcos and M. Sukhareva. Linking Etymological Databases. A Case Study in Germanic.
In 3rd Workshop on Linked Data in Linguistics: Multilingual Knowledge Resources and Natural
Language Processing, page 41, 2014.
P. Cimiano, J.P. McCrae, and P. Buitelaar. Lexicon model for ontologies: community
report, 10 May 2016. Ontology-Lexicon Community Group under the W3C Community Final
Specification Agreement (FSA), 2016. URL:
R. Cyganiak, D. Wood, and M. Lanthaler. RDF 1.1. concepts and abstract syntax:
W3C recommendation 25 February 2014, 2014. URL:
43 [15-03-2019].
LDK 2019
4:14 Shortcomings of Language Tags for Lesser-Known Languages
Gerard de Melo. Language-Related Information for the Linguistic Linked Data
Cloud. Semantic Web, 6(4):393–400, August 2015.
Th. Declerck, E. Wandl-Vogt, and K. Mörth. Towards a Pan European Lexicography by
Means of Linked (Open) Data. In I. Kosem et. al., editor, Electronic Lexicography in the 21st
Century: Linking Lexical Data in the Digital Age. Proceedings of the eLex 2015 Conference,
11-13 August 2015, Herstmonceux Castle, United Kingdom, pages 342–355. Trojina, Institute
for Applied Slovene Studies/Lexical Computing Ltd., 2015.
International Organization for Standardization. Language codes – ISO 639. URL:
F. Gillis-Webber. Conversion of the English-Xhosa Dictionary for Nurses to a linguistic linked
data framework. Information, 9(11), 2018. doi:10.3390/info9110274.
F. Gillis-Webber, S. Tittel, and C. M. Keet. A Model for Language Annotations on the Web,
2019. (submitted).
J. Gracia, M. Villegas, A. Gómez-Pérez, and N. Bel. The Apertium Bilingual Dictionaries on
the Web of Data. In Semantic Web – Interoperability, Usability, Applicability, pages 1–10. IOS
Press, 2017.
R. Güldermann. Towards casting a wider net over N
ng: chances and challenges of
archival Khoisan resources, 2014. URL:
H. Hammarström, R. Forkel, and M. Haspelmath. Glottolog 3.3., 2018. accesssed 21-02-2019.
SIL International. ISO 639-3: Relationship between ISO 639-3 and the other parts of ISO 639,
2017. URL:
SIL International. ISO 639-3: Scope of denotation for language identifiers, 2017. URL:
SIL International. ISO 639-3: Types of individual languages, 2017. URL:
R. Ishida. Language Tags in HTML and XML, 2014. URL:
F. Khan, J.E. Díaz-Vera, and M. Monachini. The Representation of an Old English Emotion
Lexicon as Linked Open Data. In John P. McCrae et al., editor, Proceedings of the LREC 2016
Workshop “LDL 2016 – 5th Workshop on Linked Data in Linguistics: Managing, Building and
Using Linked Language Resources”, 24 May 2016 – Portorož, Slovenia, pages 73–76, 2016.
23 G. Köbler. Wörterbuch des althochdeutschen Sprachschatzes. Schöningh, Paderborn, 1993.
L. Lezcano, S. Sánchez-Alonso, and A. Roa-Valverde. A Survey on the Exchange of Linguistic
Resources. Program, 47,3:263–281, 2013.
J. Lieberman, R. Singh, and Ch. Goad. W3C geospatial vocabulary: W3C incubator group
report 23 October 2007, 2007.
J.P. McCrae, J. Bosque-Gil, J. Gracia, P. Buitelaar, and P. Cimiano. The OntoLex-Lemon
model: Development and Applications. In Proceedings of ELEX 2017: Lexicography from
Scratch. September 2017, pages 19–21, 2017.
S. Moran and M. Brümmer. Lemon-aid: Using Lemon to Aid Quantitative Historical Linguistic
Analysis. In Ch. Chiarcos et al., editor, Proceedings of the 2nd Workshop on Linked Data in
Linguistics (LDL-2013), Pisa, September 2013, pages 28–33. Ass. for Comp. Linguistics, 2013.
28 A. Phillips and M. Davis. Tags for Identifiying Languages. BCP, 47, 2009.
C.M. Schlebusch, P. Skoglund, and P. Sjödin et al. Genomic Variation in Seven Khoe-San
Groups Reveals Adaptation and Complex African History. Science, 338(6105):374–379, 2012.
S. Shah and M. Brenzinger. Ouma Geelmeid ke kx’u
xa N
uu. Centre for African Language
Diversity, University of Cape Town, Cape Town, 2016.
S. Tittel, H. Bermúdez-Sabel, and Ch. Chiarcos. Using RDFa to Link Text and Dictionary Data
for Medieval French. In J.P. McCrae et al., editor, Proceedings of the Eleventh International
Conference on Language Resources and Evaluation (LREC 2018). 6th Workshop on Linked
Data in Linguistics (LDL-2018), Miyazaki, Japan, 2018, pages 30–38, Paris (ELRA), 2018.
F. Gillis-Webber and S. Tittel 4:15
S. Tittel and Ch. Chiarcos. Historical Lexicography of Old French and Linked Open Data:
Transforming the Resources of the Dictionnaire étymologique de l’ancien français with OntoLex-
Lemon. In Proceedings of the Eleventh International Conference on Language Resources and
Evaluation (LREC 2018). GLOBALEX Workshop (GLOBALEX-2018), Miyazaki, Japan,
2018, pages 58–66, Paris (ELRA), 2018.
M. Van Der Merwe. Giving breath to a dying history, 2015. URL:
W. von Wartburg. Französisches Etymologisches Wörterbuch. Eine darstellung des galloro-
manischen sprachschatzes – FEW. ATILF, since 1922. [Continued by O. Jänicke, C.T. Gossen,
J.-P. Chambon, J.-P. Chauveau, and Yan Greub].
D. Wood, M. Zaidman, L. Ruth, and M. Hausenblas. Linked data: structured data on the web.
Manning Publications Co., New York, 2014.
A Old French dialects
Table 3 List of Old French dialects (described in French) registered by the DEAF.
Abbrev. Language Abbrev. Language
afr. ancien français saint. saintongeais
mfr. moyen français tour. tourangeau
fr. du 16es. français du 16esiècle orl. orléanais
fr.dial. français dialectal bourb. bourbonnais
frc. francien (français de l’Ile de France) bourg. bourguignon
pic. picard lyon. lyonnais
flandr. français de la Flandre française frcomt. franc-comtois
hain. hennuyer francoit. franco-italien
art. artésien Nord-Est
wall. wallon Nord
liég. liégeois Nord-Ouest
champ. champenois Ouest
lorr. lorrain Sud-Ouest
norm. normand Centre
agn. anglo-normand Est
hbret. haut-breton Sud-Est
ang. angevin Terre Sainte
poit. poitevin judéofr. judéofrançais
LDK 2019
... British English written on the isle of Guernsey, for example, could be either en-GG (for geographical and political reasons, as Guernsey is not part of the United Kingdom), or en-GB (for linguistic reasons). For a deeper discussion of technical and linguistic shortcomings of language tags see [24]. ...
... As a replacement for ISOcat, Schuurman et al. [33] introduced the CLARIN Concept Registry (CCR), 24 which is based around the OpenSKOS software [35]. They aimed to avoid the issues with ISOcat by allowing only CLARIN National Content coordinators to update the registry, and by requiring a "good definition" of 142 8 Linguistic Categories a concept that is unique, meaningful, reusable and concise. ...
... We compared both approaches and formulated the recommendation to apply IETF language tags whenever possible and adequate, but to use standardized properties and repositories such as Glottolog where these are not sufficient. Gillis and Tittel [24] provide a deeper discussion of language tags and their shortcomings. It is possible that in the longer future, IETF language tags and URI-based means of language identification may converge, a recent discussion about such possibilities can be found in the context of the 'EasierRDF' initiative. ...
The (re-)usability of NLP tools and language resources has long been recognized as a key challenge in the language resource and NLP communities. Reuse of resources, however, requires a minimum level of interoperability, and in this chapter, we focus on conceptual interoperability, i.e. harmonization between different annotation schemas by means of terminology repositories. Beyond that, we give special attention to language identifiers, as these can be provided in different ways in an RDF context, either by reference to a concepts in a terminology repository, or by means of language tags.
... For example, it is European consensus that territorial varieties of languages need to be valorized and promoted, particularly online. International organizations emphasize the need for (culturally and) linguistically diverse local content to be published online; for a vitalization of multilingualism on the internet, see [30,[13][14][15][16][17][18][19][20][21]. Consequently, a fast-growing number of language resources have to be annotated, managed, and retrieved, not just for the few globally spoken languages, but also for the many local and regional languages. ...
... Within this system, a number of varieties-also called lects-reflect diatopic aspects referring to geographic areas (regional varieties, dialects, patois), diaphasic aspects referring to the communicative context (formal or informal style, technical language), and diastratic aspects referring to the social classes (sociolect, idiolect, youth language) [9]; cp. [4,14]. These thus resulted in the most relevant classes in MoLA. ...
... MoLA enables a modeler or annotator to define both periods and regions for a languoid, reflecting its diachronicity. A custom language tag, encoded using a pattern [14], can then be associated with languoid and period or region or both, which is more comprehensive than the standard ISO 639 language codes. Not only does MoLA provide dereferenceable URIs with persistent identifiers, it can also be queried, returning useful information about that languoid. ...
Several annotation models have been proposed to enable a multilingual Semantic Web. Such models hone in on the word and its morphology and assume the language tag and URI comes from external resources. These resources, such as ISO 639 and Glottolog, have limited coverage of the world’s languages and have a very limited thesaurus-like structure at best, which hampers language annotation, hence constraining research in Digital Humanities and other fields. To resolve this ‘outsourced’ task of the current models, we developed a model for representing information about languages, the Model for Language Annotation (MoLA), such that basic language information can be recorded consistently and therewith queried and analyzed as well. This includes the various types of languages, families, and the relations among them. MoLA is formalized in OWL so that it can integrate with Linguistic Linked Data resources. Sufficient coverage of MoLA is demonstrated with the use case of French.
... Consider the sentence "Stay, they said." 119 The Stanford PCFG parser 120 analyzes Stay as a verb phrase contained in (and only constituent of) a sentence. In NIF, both would be conflated. ...
... lang-subtags-templates.xhtml 166 Cf. . very notion of language tags has been criticised as being both too inflexible as well as unable to address the needs of linguistics, e.g., recently by [120,121], and alternatives are being explored [122]. URI-based language identification represents a natural alternative in such cases, as these are not tied to any single standardization body or maintainer, but allow the marking of both the respective organization or maintainer of the resource (as part of the namespace) and the individual language (in the local name). ...
Full-text available
This article provides a comprehensive and up-to-date survey of models and vocabularies for creating linguistic linked data (LLD) focusing on the latest developments in the area and both building upon and complementing previous works covering similar territory. The article begins with an overview of some recent trends which have had a significant impact on linked data models and vocabularies. Next, we give a general overview of existing vocabularies and models for different categories of LLD resource. After which we look at some of the latest developments in community standards and initiatives including descriptions of recent work on the OntoLex-Lemon model, a survey of recent initiatives in linguistic annotation and LLD, and a discussion of the LLD metadata vocabularies META-SHARE and lime. In the next part of the paper, we focus on the influence of projects on LLD models and vocabularies, starting with a general survey of relevant projects, before dedicating individual sections to a number of recent projects and their impact on LLD vocabularies and models. Finally, in the conclusion, we look ahead at some future challenges for LLD models and vocabularies. The appendix to the paper consists of a brief introduction to the OntoLex-Lemon model.
Full-text available
The English-Xhosa Dictionary for Nurses (EXDN) is a bilingual, unidirectional printed dictionary in the public domain, with English and isiXhosa as the language pair. By extending the digitisation efforts of EXDN from a human-readable digital object to a machine-readable state, using Resource Description Framework (RDF) as the data model, semantically interoperable structured data can be created, thus enabling EXDN’s data to be reused, aggregated and integrated with other language resources, where it can serve as a potential aid in the development of future language resources for isiXhosa, an under-resourced language in South Africa. The methodological guidelines for the construction of a Linguistic Linked Data framework (LLDF) for a lexicographic resource, as applied to EXDN, are described, where an LLDF can be defined as a framework: (1) which describes data in RDF, (2) using a model designed for the representation of linguistic information, (3) which adheres to Linked Data principles, and (4) which supports versioning, allowing for change. The result is a bidirectional lexicographic resource, previously bounded and static, now unbounded and evolving, with the ability to extend to multilingualism.
Conference Paper
Full-text available
This paper presents an endeavor to transform a scholarly text edition (of a medical treatise written in Middle French) into a digital edition enriched with references to an on-line dictionary. Hitherto published as a book, the resulting digital edition will use RDFa to interlink its vocabulary with the corresponding lexical entries of the Dictionnaire étymologique de l’ancien français (DEAF). We demonstrate the feasibility of RDFa for the semantic enrichment of digital editions within the philologies. In particular, the technological support for RDFa excels beyond domain-specific solutions favored by the TEI community. Our findings may thus contribute to future technological bridges between TEI/XML and (Linguistic) Linked Open Data resources. The original data of the edition is available in a L A TEX format that includes profound semantic markup. We convert this data into XML/TEI, and integrate RDFa-compliant attributes for every lexeme attested in the text. The HTML5 edition generated from the XML sources preserves the RDFa attributes and thus (a) embeds (links) its vocabulary within the overall system of the medieval French language, and that (b) provides and displays linguistic features (say, sense definitions given in the original corpus data) along with the critical apparatus of the original book publication.
Conference Paper
Full-text available
The adaptation of novel techniques and standards in computational lexicography is taking place at an accelerating pace, as manifested by recent extensions beyond the traditional XML-based paradigm of electronic publication. One important area of activity in this regard is the transformation of lexicographic resources into (Linguistic) Linked Open Data ([L]LOD), and the application of the OntoLex-Lemon vocabulary to electronic editions of dictionaries. At the moment, however, these activities focus on machine-readable dictionaries, natural language processing and modern languages and found only limited resonance in philology in general and in historical language stages in particular. This paper presents an endeavor to transform the resources of a comprehensive dictionary of Old French into LOD using OntoLex-Lemon and it sketches the difficulties of modeling particular aspects that are due to the medieval stage of the language.
Full-text available
This article illustrates the progresses made in representing a multilingual and multi-alphabetical Old Occitan medico-botanical lexicon in the context of the project Dictionnaire de Termes Médico-botaniques de l'Ancien Occitan (DiTMAO). The chosen lexical model of reference is lemon, which has been extended accordingly to some specific linguistic and lexical features of the lexicon. In particular, issues and solutions about the modeling of multilingual and multiword phenomena are discussed, as the way they are managed through LexO, a web editor developed in the context of the project.
Full-text available
‘Open Data’ has become very important in a wide range of fields. However for linguistics, much data is still published in proprietary, closed formats and is not made available on the web. We propose the use of linked data principles to enable language resources to be published and interlinked openly on the web, and we describe the application of this paradigm to the modeling of two resources, WordNet and the MASC corpus. Here, WordNet and the MASC corpus serve as representative examples for two major classes of linguistic resources, lexical-semantic resources and annotated corpora, respectively.Furthermore, we argue that modeling and publishing language resources as linked data offers crucial advantages as compared to existing formalisms. In particular, it is explained how this can enhance the interoperability and the integration of linguistic resources. Further benefits of this approach include unambiguous identifiability of elements of linguistic description, the creation of dynamic, but unambiguous links between different resources, the possibility to query across distributed resources, and the availability of a mature technological infrastructure. Finally, recent community activities are described.
Full-text available brings information about languages, words, and other linguistic entities to the Web of Linked Data. It defines URIs for terms, languages, scripts, and characters, which are not only highly interconnected but also linked to a variety of resources on the Web. Additionally, new datasets are being published to contribute to the emerging Linked Data Cloud of Language-Related information.
Full-text available
The term Linked Data refers to a set of best practices for publishing and connecting structured data on the Web. These best practices have been adopted by an increasing number of data providers over the last three years, leading to the creation of a global data space containing billions of assertions-the Web of Data. In this article we present the concept and technical principles of Linked Data, and situate these within the broader context of related technological developments. We describe progress to date in publishing Linked Data on the Web, review applications that have been developed to exploit the Web of Data, and map out a research agenda for the Linked Data community as it moves forward.
Several annotation models have been proposed to enable a multilingual Semantic Web. Such models hone in on the word and its morphology and assume the language tag and URI comes from external resources. These resources, such as ISO 639 and Glottolog, have limited coverage of the world’s languages and have a very limited thesaurus-like structure at best, which hampers language annotation, hence constraining research in Digital Humanities and other fields. To resolve this ‘outsourced’ task of the current models, we developed a model for representing information about languages, the Model for Language Annotation (MoLA), such that basic language information can be recorded consistently and therewith queried and analyzed as well. This includes the various types of languages, families, and the relations among them. MoLA is formalized in OWL so that it can integrate with Linguistic Linked Data resources. Sufficient coverage of MoLA is demonstrated with the use case of French.
Linked Data presents the Linked Data model in plain, jargon-free language to Web developers. Avoiding the overly academic terminology of the Semantic Web, this new book presents practical techniques using everyday tools like JavaScript and Python.
Bilingual electronic dictionaries contain collections of lexical entries in two languages, with explicitly declared translation relations between such entries. Nevertheless, they are typically developed in isolation, in their own formats and accessible through proprietary APIs. In this paper we propose the use of Semantic Web techniques to make translations available on the Web to be consumed by other semantic enabled resources in a direct manner, based on standard languages and query means. In particular, we describe the conversion of the Apertium family of bilingual dictionaries and lexicons into RDF (Resource Description Framework) and how their data have been made accessible on the Web as linked data. As a result, all the converted dictionaries (many of them covering under-resourced languages) are connected among them and can be easily traversed from one to another to obtain, for instance, translations between language pairs not originally connected in any of the original dictionaries.