Conference PaperPDF Available

Identification of Languages in Linked Data: A Diachronic-Diatopic Case Study of French

Authors:

Abstract and Figures

When modelling linguistic resources as Linked Data, the identification of languages using language tags and language codes is a mandatory task. IETF's BCP 47 defines the standard for tags, and ISO 639 provides the codes. However, these codes are insufficient for the identification of diatopic variation within a language and, also, for different historical language stages. This weakness hampers the accurate identification of data, which in turn leads to ambiguity when extending, aggregating and re-using this data-a key notion of Linked Open Data and the Semantic Web. We show the limitations of language identification with a case study of French linguistic data from both a diachronic and a diatopic perspective. Our exemplary data derives from dictionaries of Old French, Middle French, and of Modern French dialects, and from a Modern French linguistic atlas. For each exemplar, we propose a solution using the privateuse sub-tag of BCP 47's language tag, staying within the boundaries of existing standards. Using a predefined pattern for the privateuse sub-tag, the solutions enable a dialect, a patois, in combination with a time period, to be defined and identified. This can lead to shared agreement of language tags that will increase interoperability within the context of Linked Data.
Content may be subject to copyright.
Identification of Languages in Linked Data:
A Diachronic-Diatopic Case Study of French
Sabine Tittel
1
, Frances Gillis-Webber
2
1
Heidelberg Academy of Sciences and Humanities, Seminarstraße 3,
D–69117 Heidelberg, Germany
2
Department of Computer Science, University of Cape Town, Cape Town, South Africa
E-mail: sabine.tittel@urz.uni-heidelberg.de, fran@fynbosch.com
Abstract
When modelling linguistic resources as Linked Data, the identification of languages using
language tags and language codes is a mandatory task. IETF’s BCP 47 defines the standard
for tags, and ISO 639 provides the codes. However, these codes are insufficient for the
identification of diatopic variation within a language and, also, for different historical language
stages. This weakness hampers the accurate identification of data, which in turn leads to
ambiguity when extending, aggregating and re-using this data—a key notion of Linked Open
Data and the Semantic Web. We show the limitations of language identification with a case
study of French linguistic data from both a diachronic and a diatopic perspective. Our
exemplary data derives from dictionaries of Old French, Middle French, and of Modern French
dialects, and from a Modern French linguistic atlas. For each exemplar, we propose a solution
using the privateuse sub-tag of BCP 47’s language tag, staying within the boundaries of existing
standards. Using a predefined pattern for the privateuse sub-tag, the solutions enable a dialect,
a patois, in combination with a time period, to be defined and identified. This can lead to
shared agreement of language tags that will increase interoperability within the context of
Linked Data.
Keywords: language codes; language tags; language annotation; Linked Open Data; French
dialects
1. Introduction
Over the last decade, modelling linguistic data using the Resource Description
Framework (RDF), following the Linked Data (LD) paradigm, has become a
widespread method for the creation of datasets for a multilingual web of data. It enables
machine-readable, cross-resource access to data that are otherwise spread across the
web as isolated datasets. However, for the modelling of linguistic resources as LD, the
use of language tags is essential: the annotation with language tags whose form adheres
to established standards ensures unambiguous language identification of linguistic
information, such as lexemes and their graphic and phonetic realizations. Because of
the interlinking of lexemes and their different realizations, the LD format can be
particularly valuable for linguistic resources that document the diatopic diversity of a
given language (i.e., with a spatial reference). Examples are regional dictionaries or
linguistic atlases. These resources can be complemented with historical data to
547
Proceedings of eLex 2019
introduce a diachronic perspective to the diatopic variation of the language (i.e.,
considering evolution through history). This can be, e.g., data from historical
dictionaries that indicate regional characteristics. The publication of these resources as
LD and the corresponding means of data query can enhance studies that focus on the
diatopic richness of modern-day languages and on the evolution of diatopic variation
at the same time. The use of language tags is specified by IETF’s BCP 47 (Phillips &
Davis, 2009: 1-4) and the required language codes come from ISO 639 (International
Organization for Standardization, n.d.). Within our field, however, we observe a lack
of language tags and codes hampering the required language annotation. In this paper,
we address the issue of language tagging with French linguistic resources combining a
diatopic with a diachronic perspective: in a case study, we investigate data of Old-,
Middle- and Modern French resources with (regional) dictionary data and linguistic
atlas data.
After a short outline of the diachronic-diatopic landscape of French linguistic resources
(Section 1.1), we briefly describe RDF, LD (Section 1.2), and the identification of
languages (Section 1.3). In the following section, we introduce the use of a pattern for
language tags (Section 2). Our case study of French uses exemplary data of historical
and modern dictionaries (Section 3) and of a linguistic atlas (Section 4). For each
exemplar, we demonstrate a solution for the language tagging, using the pattern
described. We evaluate the solutions in Section 5, and in Section 6, we present an
interface which can be used to generate (and decode) language tags according to our
pattern. We conclude the paper in Section 7.
1.1 Diatopic linguistic resources and a diachronic perspective
The regional varieties, dialects and patois
1
of the French of France are under-
represented in linguistic consideration in general and in lexicography in particular
(Rézeau, 2001: 7). This is all the more true for the diatopic reflection from a diachronic
angle: the historical development of French regionalisms has not been studied in a
comprehensive yet detailed way (Gleßgen & Thibaut, 2005: XII). Studies focusing on
single topics such as a particular region in a particular time period have been
conducted, recently by, e.g., Chauveau (2016), and Rézeau (2016).
There are many resources that can be exploited for diatopic-diachronic studies: for the
different language periods of French, dictionaries, corpora, and—for modern French in
1
We are aware of the discussion of the terms that denote different variations within the
diatopic diasystem of French. In this paper, we will use the terms following the French
literature, where régionalité linguistique (of French) is clearly distinguished from dialectes,
the first referring to variation within the standard language, the latter to the primary
dialects of France that are the successors of the Old French dialects (Gleßgen & Thibaut,
2005: V), and patois typically designating a local variety of a dialect. Note that we use
‘patois’ as a non-pejorative term.
548
Proceedings of eLex 2019
particular—linguistic atlases are available.
2
Modern resources covering French varieties
include dialect or patois dictionaries (e.g., Rézeau, 2001; Varlet, 1896; Vasseur, 1998),
linguistic atlases (e.g., Gilliéron & Edmont, 1902–1910; Lanher et al., 1979–1988;
Dondaine & Dondaine, 1972–1991), corpora (Thun, 2011)
3
, and, also, individual studies
(e.g., Rézeau, 2007) focusing on regional French, dialects and patois. For the historical
language stages however, there are fewer resources with diatopic content. A reason for
this is that from ca. 1500 AD—with the constitution of French (evolving from a
Parisian scripta
4
that had occurred around 1250) as a national language (Wolf, 1979:
94f.)—to the beginning of the 19
th
century, dialects almost exclusively belonged to the
oral culture (Berschin et al., 2008: 203–211). Consequently, studies on the subject of
regionalisms are scarce for this time period. Earlier however, in medieval times, the
primary dialects included in the notion of Old- and Middle French, such as Picard and
Anglo-Norman, were used for both oral and written communication. Hence, we look at
the transmission of numerous linguistic primary resources (texts in manuscripts, often
accessible in scholarly text editions) documenting regional variation during the Middle
Ages. For this time period, studies mainly focus on a single primary resource and how
to localize its language in a specific region (notably works by J.-P. Chambon, e.g.,
Chambon, 1997, and G. Roques, cf. the ‘Liste Roques’ in Glessgen & Trotter, 2016:
473–635). There are also many-volumed, comprehensive dictionaries of the historical
language stages, in particular the Dictionnaire étymologique de l’ancien français
(DEAF, Baldinger et al., 1971–) for Old French, the Dictionnaire du moyen français
(DMF, ATILF – CNRS & Université de Lorraine (2015)) for Middle French, and the
Französisches Etymologisches Wörterbuch (FEW, von Wartburg, 1922–) for the
diachronic description of French until the present day. These dictionaries—although
not necessarily conceived as data sources for diatopic linguistics—provide a synopsis of
the knowledge of the particular historical language stage. By incorporating the results
of historical dialect studies, they thus contribute to our knowledge of regional variation
evolving through time.
Digitization of diatopic resources. It is a European consensus that geographic variation
of languages needs to be valorized and promoted, particularly online: UNESCO, La
Francophonie
5
and other international organizations emphasize the need for (culturally
and) linguistically diverse local content to be published online and for a vitalization of
multilingualism on the Web, cf. Vannini & Le Crosnier, 2012: 13–21. A large number
of the resources in our focus—word lists, dictionaries, linguistic atlases, texts—are
currently only available in print. Only a few are available in digital form, and mostly
2
We identified five language periods of French, cf. Gillis-Webber et al. (2019: Section 4 with
Fig. 4).
3
Corpus of letters written by prisoners, soldiers, prostitutes, etc., that document the diatopic
variation within the French substandard language.
4
The written form of a spoken dialect.
5
https://www.unesco.com/; https://www.francophonie.org/ [13-02-2019].
549
Proceedings of eLex 2019
as digital images.
6
Many have yet to be (retro-)digitized. Digitization would allow for
“many new approaches to the quantitative comparison of languages, be it for a better
understanding of cross-linguistic variation in grammatical structure or for new and
improved historical comparative reconstructions” (Bouda & Cysouw, 2012: 15). One
such approach is the representation of the resource in RDF, which in turn allows for
the extension to LD.
1.2 Enabling resource integration with the Resource Description
Framework and Linked Data
RDF
7
is a data model that represents knowledge in a graph data structure facilitating
data interchange on the (Semantic) Web. It is a fundamental technology of the
Semantic Web, in which data is structured and meaning can thus be inferred by
machines. RDF expresses data as sets of statements in the form of subject-predicate-
object-triples. Each subject and object is a node; the predicate (or property) forms a
relation (edge) pointing from the source node (subject) to a target node (object). Nodes
and edges are identified with URIs (Uniform Resource Identifier, accessible via HTTP),
and the object can also be described as a string literal (Cyganiak et al., 2014). LD can
be described as a set of recommended practices for publishing RDF as structured data
on the Web (Bizer et al., 2009). Applying LD principles (Berners-Lee, 2006) to the
modelling of linguistic data comes with significant advantages, such as structural
interoperability (cross-resource access by using same format and same query language),
conceptual interoperability (through shared vocabularies), accessibility (through
standard Web protocols), and resource integration by means of interlinking (Chiarcos
et al., 2013). Because of the exploratory nature of LD, URIs identifying, e.g., lexemes,
their senses, and their concepts referring to the things denoted, things and the usage
of their designations can be explored in a cultural context without being restricted to
the vehicle of a particular language. The integration of resources of different language
stages and diatopic variation enables observation through time and space, including,
e.g., borrowing and word formation processes, and semantic shift within a large data
collection. For Old French, the first steps have been made by modelling exemplary
lexicographic data of the DEAF as LD using the OntoLex-Lemon vocabulary
8
, and the
modelling of a scholarly text edition of a Middle French medical treatise using RDFa
(Tittel & Chiarcos, 2018; Tittel et al., 2018). To the best of our knowledge, there are
no other historical linguistic resources of French represented as LD that could be
exploited for diachronic-diatopic studies.
6
Cf., e.g., the references at https://www.lexilogos.com/lorrain_dictionnaire.htm [10-06-2019].
7
RDF 1.1. Primer, 2014, https://www.w3.org/TR/rdf11-primer/ [10-05-2019].
8
https://www.w3.org/2016/05/ontolex/ [13-05-2019].
550
Proceedings of eLex 2019
1.3 Identification of languages
When modelling linguistic resources in RDF, it is necessary to identify the language of
the resource and the information therein (be it a word, a multiword expression, a sense,
a graphical realization, a phonetic representation), and to annotate literals with a
language tag. IETF’s BCP 47 specifies the Best Current Practice for language tags; the
language tag typically begins with a language code and it must conform to established
standards (Cyganiak et al., 2014). The language code comes from external resources
such as ISO 639, which provides the authoritative list of language codes. Alternatives
are catalogues like Glottolog, Ethnologue, and MultiTree.
9
However, these alternatives
do not meet the requirements of BCP 47 for the encoding of languages. They also reveal
significant shortcomings concerning registration, hierarchization, diachronic and
dialectal criteria, all of which have been discussed in detail in Gillis-Webber and Tittel
(2019: 4:6-8) and Gillis-Webber et al. (2019). Lexvo
10
provides dereferenceable URIs
only for languages registered by ISO 639 (de Melo, 2015). It is, thus, insufficient for
our use.
An exemplary lexical entry in RDF (identified as E0), modelled using OntoLex-Lemon
and serialized in Turtle
11
is:
1 @PREFIX : <http://www.example.com/entry/> .
2 @PREFIX ontolex: <http://www.w3.org/ns/lemon/ontolex#> .
3 @PREFIX lexinfo: <http://www.lexinfo.net/ontology/2.0/lexinfo#> .
4 @PREFIX dct: <http://purl.org/dc/terms/> .
5 @PREFIX rdfs: <http://www.w3.org/2001/02/rdf-schema#> .
6 @PREFIX dbpedia: <http://www.dbpedia.org/resource/> .
7
8 :alconorque a ontolex:LexicalEntry , ontolex:Word ;
9 lexinfo:partOfSpeech lexinfo:Noun ;
10 dct:language <http://lexvo.org/id/iso639-1/pt> ,
11 <https://iso639-3.sil.org/code/por> ;
12 rdfs:label "cork oak"@en , "alconorque"@pt ;
13 ontolex:denotes dbpedia:Quercus_suber .
9
https://glottolog.org, https://www.ethnologue.com, http://multitree.org/ [07-06-2019].
10
http://lexvo.org [07-06-2019].
11
Terse RDF Triple Language, http://www.w3.org/TR/turtle/ [11-01-2019]. In the following
code examples, namespaces are assumed defined the usual way. We include hypothetical
URIs using the namespace <http://www.example.com/entry/>.
551
Proceedings of eLex 2019
where Lines 10-11 show the applicable language URIs for the lexical entry indicated as
‘Portuguese’, from ISO 639-1 and ISO 639-3 respectively; Line 12 shows the language
code ‘English’ (ISO 639-1 ‘en’) for the literal “cork oak”, and the language code
‘Portuguese’ (ISO 639-1 ‘pt’) for the literal “alconorque”.
The ISO 639 standard shows significant shortcomings with respect to regional variation
and to historical language stages, as was shown in Gillis-Webber and Tittel (2019: 4:4-
5); cf. also Figures. 4 and 5. This prevents the unambiguous identification of languages,
even more so when modelling multiple ‘snapshots’ of data of the same language through
time and space.
2. Pattern for Language Tags
As demonstrated in E0, the use of ISO 639 language codes in language tags is
straightforward for most modern and well-known languages. However, the problem of
missing or inadequate language codes extends to any variety or dialect of a language
which is requires representation on the web, and for which an ISO 639 code is simply
not available. Language tags, as prescribed by BCP 47, have the syntax:
language-extlang-script-region-variant-extension-privateuse
with each portion, called a sub-tag, separated by a hyphen (Phillips & Davis, 2009: 4).
Gillis-Webber & Tittel (2019) propose a pattern for the privateuse sub-tag.
12
The
pattern for the privateuse sub-tag is of the form:
x-language-otherlect-timeperiod-region-uri
where x- is a BCP 47 requirement indicating privateuse, and language (a language,
dialect, patois or pidgin), otherlect (an ethnolect, sociolect, or idiolect), timeperiod,
region, and URI are all parts of the sub-tag, separated by a hyphen (Gillis-Webber &
Tittel, 2019: 4:12). Apart from the privateuse sub-tag, the sub-tags are specified by
BCP 47 as “identified on the basis of its length, position in the tag, and its content”;
each sub-tag typically is part of an ISO standard or registry (ib.) For the privateuse
sub-tag, the use of a key (Table 1) is proposed to identify each part, thus allowing for
flexibility of content and variable length thereof.
12
Note that this pattern is not intended to replace any content that would typically be
included in other sub-tags. To see the most recent updates to the pattern, please go to:
https://londisizwe.org/ language-tags/.
552
Proceedings of eLex 2019
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
4
= Identifier from GeoNames
URI 4 0 = URI shortcode from https://londisizwe.org/language-tags/
Table 1: The key for each part in the privateuse tag.
We identified the following set of competency questions (CQs) for the pattern, where
[lect] can be replaced by any language, variant, dialect, patois, and scripta.
CQ 1 How to identify a [lect] that has no ISO 639 language code, but whose parent
language does?
CQ 2 How to identify a [lect] for which ISO 639 provides a language code that indicates
a different time period?
CQ 3 How to identify a [lect] for which ISO 639 provides two language codes?
CQ 4 How to identify a [lect] in space that has neither an ISO 639 code nor a code
from an alternative directory?
CQ 5 How to identify a [lect] in time?
CQ 6 How to identify endonyms and exonyms of a [lect]?
When evaluating the pattern, these CQs should be answerable. Using the case study
of French, we will revisit the CQs in Section 5 to test the efficacy of the proposed
pattern.
553
Proceedings of eLex 2019
3. Modelling of regional variation in dictionary data
For our case study, we will embrace both diachronic and diatopic data of French, with
the latter typically mirroring aspects of the former.
3.1 Old French
Old French should be understood as an umbrella term for a number of dialects resulting
from the process of settlement and romanization, different substrates, strates, etc.
These dialects present distinctive linguistic realities from the beginning of the 12
th
century, cf. Rickard (1974: 54–65; 71–84).
For the Old French period, the contribution of the DEAF to our knowledge of diatopic
variation of Old French has been discussed by Möhren (2016) and Tittel (2016). The
DEAF allows for the annotation of data with 35 scriptae, including broader categories
like ‘Nord-Est’ or ‘Centre’ (cf. Figures 4 and 5). For Old French, the ISO 639-3
language code is ‘fro’ («842–ca.1400»), but there are no ISO 639 language codes
available for the scriptae except for Anglo-Norman (‘xno’) and Judéo-French (‘zrp’).
For the modelling of DEAF data with OntoLex-Lemon, although ‘fro’ has been used
as the language tag, this does not allow for the data to be differentiated on scriptae
(Tittel & Chiarcos, 2018: 64f.).
An exemplar (E1) derived from the DEAF is jannaie (designating a terrain covered
with gorse), a lexeme marked as Gallo.
13
It can be modelled as follows:
1 :jannaie a ontolex:LexicalEntry , ontolex:Word;
2 ontolex:canonicalForm :jannaie_lemma .
3
4 :jannaie_lemma a ontolex:Form ;
5 ontolex:writtenRep "jannaie"@fro-x-00gallo .
In our language tag on Line 5, as an ISO 639 language code does not exist for (Old)
Gallo, we have made use of a compiled language tag: fro identifies it as from the Old
French period, and 00 indicates that it is a user-defined language (i.e., a code from an
alternative directory to ISO 639 has not been used).
14
13
DEAF J 136,9; https://deaf-server.adw.uni-heidelberg.de/lemme/jaon#jannaie [10-05-
2019].
14
For a discussion of further approaches to language tagging Old French dialects, cf. Gillis-
Webber & Tittel (2019: 4:9-11).
554
Proceedings of eLex 2019
3.2 Middle French
The comprehensive dictionary for the Middle French period is the DMF. With respect
to the study of dialectal characteristics of the Middle French lexis, the DMF is a
resource of limited value and difficult access (Renders, 2016: 95f.). However, the DMF
has the potential for facilitating the study of diatopic variation of late medieval French:
the data structure of the DMF entry does not contain a label that specifically tags
information as being dialectal (thus, the information cannot easily be accessed in a
machine-aided way), but the running (unstructured) text of approx. 1,190 entries
(Renders, 2016: 89) includes in effect such information; this can be exploited.
Although the French written standard spread in Middle French time, the dialects still
maintained their role in the literature. The DMF defines a list of 29 “étiquettes
régionales” (Renders, 2016: 86) comparable with the DEAF scriptae list. For Middle
French, the ISO 639-3 language code is ‘frm’ («ca. 1400–1600»); this can be utilized to
identify the language, but the challenge of codes for its dialects needs to be addressed.
In the following exemplar (E2), we model a lexeme that is marked as dialectal: appreper
v. “s’approcher (d’un lieu)” “Région. (Wallonie)”.
15
The language code from ISO 693-1
for modern Walloon is ‘wa’. But as for the Old French language period, the code should
not be used for the Middle French period.
1 :appreper a ontolex:LexicalEntry , ontolex:Word ;
2 ontolex:canonicalForm :appreper_lemma .
3
4 :appreper_lemma a ontolex:Form ;
5 ontolex:writtenRep "appreper"@frm-x-00walloon .
In our language tag on Line 5, frm identifies it as from the Middle French period, with
00 indicating that it is a user-defined language (cp. E1).
3.3 Modern French
Today, standard French is dominant in all regions of France. Nevertheless, regional
variation, dialects and patois characterize its linguistic landscape (Wolf, 1979: 165).
This is illustrated, e.g., by the many dictionaries and surveys referenced by Lexilogos
for French dialects. Attempts to revive regional varieties gave impetus to the creation
of many linguistic atlases of France, beginning as early as 1897-1901 with the Atlas
linguistique de la France – ALF (Gilliéron & Edmont, 1902–1910, Fig. 1a) and leading
15
http://atilf.fr/dmf/definition/appreper [01-03-2019].
555
Proceedings of eLex 2019
to the many large-sized volumes of the series Atlas linguistiques de la France par
régions – ALFR (Séguy, 1973: 78).
The language code for Modern French is ISO 639-1 ‘fr’. For the majority of French
regional varieties, ISO 639 codes are not available, exceptions being ISO 639-3 ‘nrf’ for
the Norman dialect
16
, ‘pcd’ for Picard, and ‘wln’ for Walloon.
Given the amount of linguistic resources with diatopic data for modern French, we
have selected exemplary data, namely from dictionaries of different patois. We focus
on one use
(a) (b)
Figure 1: (a) ALF map n
o
668 ‘grenouille’. (b) Denizot (1910: 120).
case: the designations for the frog. To model the data simply using ‘fr’ as the language
code does not account for the linguistic reality in the regions in our focus: it would
render the diatopic variation generic. BCP 47 specifies a region sub-tag that is typically
used to indicate (diatopic or diastratic) variation within a country or territory, the
standard being a code from ISO 3166. However, ISO 3166 registers administrative (sub-
)divisions (in our case, régions and départements of contemporary France) whose
boundaries do not necessarily match the language boundaries.
17
Hence, we make use of
the privateuse subtag and codes provided by Glottolog, e.g., for Burgundian in E3
(‘bourg1247’), in line with the pattern in Table 1. However, the patois spoken in
Burgundy (and in any other region) differ. It is thus necessary to further distinguish
16
Falsely described as “Guernésiais, Jèrriais” which excludes the continental area.
17
https://tools.ietf.org/html/bcp47#section-2.2.4;
https://www.iso.org/obp/ui/#iso:code:3166:FR [1106-2019].
556
Proceedings of eLex 2019
the language tag on patois. We do this by adding the name of the location where the
patois has been recorded. This can be (1) a region or (2) a place name.
To identify a language in a region (1), as a subset of the language denoted by the
Glottocode, we use the latitude and longitude coordinates of the location provided by
the geographical database GeoNames
18
and we convert the coordinates into a Geohash
19
,
where Geohash is a system for encoding geographic coordinates as a base32 string, in
a syntax acceptable for BCP 47 (Gillis-Webber & Tittel, 2019: 4:10). To identify a
place name (2) within the language tag, we refer to its equivalent entry in GeoNames.
3.3.1 Language of Burgundy
E3, from Dictionnaire de patois de Mancey (Millot (1905–1922 (edition 1998)):
1 @PREFIX pwn: <http://wordnet-rdf.princeton.edu/id/> .
2
3 :gornaille a ontolex:LexicalEntry , ontolex:Word ;
4 :rdfs:label "gornaïlle"@fr-x-01bour1247-342996271 ;
5 ontolex:canonicalForm :gornaille_lemma ;
6 ontolex:sense :gornaille_sense ;
7 ontolex:evokes :frog_lexConcept.
8
9 :gornaille_lemma a ontolex:Form ;
10 ontolex:writtenRep "gornaïlle"@fr-x-01bour1247-342996271 .
11
12 :gornaille_sense a ontolex:LexicalSense ;
13 ontolex:isLexicalizedSenseOf :frog_lexConcept .
14
15 :frog_lexConcept a ontolex:LexicalConcept ;
16 ontolex:lexicalizedSense :gornaille_sense ;
17 ontolex:isConceptOf dbpedia:Frog ;
18 ontolex:definition "grenouille"@fr ;
19 dct:references pwn:01642406-n .
In our language tag on Lines 4 and 10, fr identifies the tag as from the Modern French
period, with 01 indicating that the Glottocode for the Burgundy language is used. To
18
https://www.geonames.org/ [07-06-2019].
19
https://www.movable-type.co.uk/scripts/geohash.html [07-06-2019].
557
Proceedings of eLex 2019
identify the patois spoken in Mancey, a commune in the Saône-et-Loire département,
we made use of the equivalent identifier from GeoNames, 2996271, prepending it with
34 as per Table 1.
E4, from the Vocabulaire patois de Sainte-Sabine et ses environs (Côte-d’Or) (Denizot
(1910), Fig. 1b):
1 :renoille a ontolex:LexicalEntry , ontolex:Word ;
2 rdfs:label "renoille"@fr-x-00saintesabine-30u0g6r--
3 u0e36--u07zp--u0sbk--u0t5k--u0u4u ;
4 ontolex:canonicalForm :renoille_lemma ;
5 ontolex:sense :renoille_sense ;
6 ontolex:evokes :frog_lexConcept .
7
8 :gueurnouille_lemma a ontolex:Form ;
9 ontolex:writtenRep "renoille"@fr-x-00saintesabine-30u0g6r--
10 u0e36--u07zp--u0sbk--u0t5k--u0u4u .
11
12 :gueurnouille_sense a ontolex:LexicalSense ;
13 ontolex:isLexicalizedSenseOf :frog_lexConcept .
The use of GeoNames to identify the location of Sainte-Sabine, a commune in the
Côted’Or département, would be a wrong approach for this case: the title of the
resource clearly indicates that the vocabulary has been recorded in Sainte-Sabine and,
also, within its vicinity. Unfortunately, the introduction of the resource gives only a
vague description of what it means: “montagnes des environs des Pouilly-en-Auxois et
de Blignysur-Ouche”, Denizot (1910: 14). We drew a polygon of the area that is, thus,
only an approximation as well (Figure 2a). The geographic coordinates representing
the polygon are: (49.62686,4.91473), (48.04287,4.66964), (47.6435,5.59192),
(47.88325,6.85844), (48.40865,7.23867), (49.72584,5.81263), (49.62686,4.91473).
The last coordinate is the same as the first, and so we excluded the last one and then
converted the latitude and longitude coordinates to a Geohash to a precision of five
digits, cf. Gillis-Webber and Tittel (2019: 4:10f.): u0g6r--u0e36--u07zp--u0sbk--u0t5k--
u0u4u. Lines 2-3 and 9-10 show the use of these Geohashes, with the pattern 00 defining
the language as user-defined and 30 defining a geohashed polygon region.
558
Proceedings of eLex 2019
(a) (b)
Figure2: (a) Approximate region where the patois of Sainte-Sabine was recorded. (b) Region
of Vimeu in Picardy
3.3.2 Language of Picardy
E5, from Dictionnaire des parlers picards du Vimeu (Somme) (Vasseur (1998)):
1 :guernouille a ontolex:LexicalEntry , ontolex:Word ;
2 rdfs:label
3 "guérnouille"@pcd-x-30u0cje--u0cj3--u0buz--u0chj--u0cm1 ;
4 ontolex:canonicalForm :guernouille_lemma ;
5 ontolex:sense :guernouille_sense ;
6 ontolex:evokes :frog_lexConcept .
7
8 :guernouille_lemma a ontolex:Form ;
9 ontolex:writtenRep
10 "guérnouille"@pcd-x-30u0cje--u0cj3--u0buz--u0chj--u0cm1 .
11
12 :guernouille_sense a ontolex:LexicalSense ;
13 ontolex:isLexicalizedSenseOf :frog_lexConcept .
In the language tag on Lines 3 and 10, the language code uses the ISO 639-3 code ‘pcd’
for the modern Picard language. To specify the region of Vimeu in Picardy (Fig. 2b),
we have again defined a region, converted into Geohashes.
559
Proceedings of eLex 2019
4. Modelling of regional variation using linguistic atlas data
We modeled a small set of exemplary data from the ALF. It seems clear to us that
most of the regional differences manifested in a linguistic atlas concern phonetic
variation. However, the regional particularities also concern the lexis, especially in
border regions of France. These regions document phenomena of cultural and linguistic
contact with other languages, e.g., with German, Franco-Provençal, Occitan, and
Breton. These phenomena are of great interest, in particular to researchers in Historical
Linguistics and Digital Humanities. With its rich lexical and phonetic data, an atlas
could add significant value to the landscape of semantically accessible linguistic data
sets.
For the transformation of linguistic atlas data into LD, the information on a map needs
to be turned into points. This leads to two issues: dealing with (a) the geographic data
acquisition points (which, in the context of ALF, is place names) and (b) the phonetic
transcription indicated for each point.
For (a), Gally et al. (2013: 188f.) describe that they semi-automatically provided each
of the 992 data acquisition points of the digitized ALF with geographic coordinates.
For (b), typically, the data sources for the linguistic atlases are surveys where
interviewees pronounced words and phrases and interviewers transcribed the phonetic
realizations using a phonetic alphabet. For the ALF, Abbé Rousselot and Jules
Gilliéron established a phonetic alphabet in 1891 which then was also used by the
makers of the atlases of the series ALFR. The transcriptions were written onto the
maps by hand. To ensure the structural interoperability of atlas data within the
Semantic Web, the transcriptions need to be re-encoded using the standard
International phonetic alphabet (IPA, International Phonetic Association, 2005), cp.
Moran (2012) who uses IPA as an interlingual pivot for different transcription systems.
4.1 Exemplary data for Lorraine
We have used data from the ALF map n
o
668 (Fig. 1a). In E6, for the lexeme grenouille
“frog”, we model the phonetic realizations of three acquisition points taken from the
Meurthe-et-Moselle département in Lorraine (Table 2) using the phoneticRep property
of the OntoLex-Lemon vocabulary.
n
o
162 (Sexey-les-Bois)
n
o
170 (Moncel-sur-Seille)
n
o
171 (Mailly-sur-Seille)
560
Proceedings of eLex 2019
Table 2: Extract from ALF map n
o
668.
E6, from Atlas linguistique de la France (Gilliéron & Edmont, 1902–1910):
1 :grenouille a ontolex:LexicalEntry , ontolex:Word ;
2 rdfs:label "grenouille"@fr ;
3 ontolex:canonicalForm :grenouille_lemma ,
4 ontolex:sense :grenouille_sense ;
5 ontolex:evokes :frog_lexConcept .
6
7 :grenouille_lemma a ontolex:Form ;
8 ontolex:writtenRep "grenouille"@fr ;
9 ontolex:phoneticRep "gK@nu–:j"@fr-fonipa-x-01lorr1242-342996683 ,
10 "g@rnu–:j"@fr-fonipa-x-01lorr1242-342974669 ,
11 "dZ@rnu–:j"@fr-fonipa-x-01lorr1242-342993415 .
12
13 :grenouille_sense a ontolex:LexicalSense ;
14 ontolex:isLexicalizedSenseOf :frog_lexConcept .
In Lines 9-11, we have re-encoded the phonetic transcription (cf. Table 2) using IPA
characters. To identify the phonetic characters of the string literals, we include the
subtag fonipa, which is compliant with BCP 47 (Phillips & Davis, 2009: 43). In the
privateuse portion, 01 indicates a code from Glottolog has been used. As with E3, the
place name for each geographic acquisition point has been represented by its equivalent
GeoNames identifier, prepended with 34. E.g., the phonetic representation of the
lexeme recorded in Sexey-les-Bois (n
o
162, Line 10) is identified as 2974669.
20
20
http://www.geonames.org/2974669/sexey-les-bois.html [06-06-2019].
561
Proceedings of eLex 2019
5. Discussion
Revisiting the CQs, all questions, with the exception of CQ6, are answerable with the
available data from our case study.
CQ1 is answered by E1E4 and E6. For E1 and E2, codes exist in alternative
directories, but they do not reflect the correct time periods. Hence, we opted to identify
the language using a user-defined code, indicated by 00 from Table 1. CQ2 is, thus,
also answered by these two exemplars. For E3, E4 and E6, a Glottocode is available,
indicated by 01 from Table 1.
CQ3 is answered by our Modern French exemplars. Although different language codes
are available for Modern French in each ISO 639 part, we make use of ‘fr’ from ISO
639-1; as per the BCP 47 specification, the shortest language code available has to be
used.
CQ4 is answered by E3E6 showing two solutions: (1) E3 and E6 make use of an
identifier from GeoNames, indicated by 34 from Table 1, (2) E4 and E5 both make
use of a user-defined language (defined with pattern 00) and of Geohashes that
represent the geographic coordinates for a polygon shaped region (defined with pattern
30 and with -- serving as an internal delimiter between each Geohash). A detailed
description of associating a geographic area with a language is discussed in Gillis-
Webber and Tittel (2019), which also addresses CQ5.
Although the pattern allows for a more precise definition of the language in question,
for E4 and E5 the language tags intuitively feel too long: the Geohash, while useful,
is opaque, and may require further annotation in order to be human-readable. While
the proposed pattern serves as an interim solution for language-tagging lesser-known
or less-discussed languages, the problem still remains that the dependency of a language
tag on an ISO standard or registry is a flaw of language tags and the RDF specification.
As an alternative to a language tag, we should be able to encode a URI in the vein of
"jannaie"@deaf:fro/gallo, where deaf: is the namespace.
Gillis-Webber and Tittel (2019) suggest exploring the creation of a sub-datatype for
rdf:langString, which would thus allow for the datatype URI to be encoded, as an
alternative to the language tag. However, doing this presents challenges. A literal
consists of two elements: a lexical form and a datatype URI (Cyganiak et al., 2014). If
the datatype URI is http://www.w3.org/1999/02/22rdf-syntax-ns#langString, then a
third element is introduced to the literal: namely “a non-empty language tag as defined
by BCP 47”, ib. All other datatype URIs are mapped to RDF-compatible XSD types,
none of which would allow the introduction of a custom URI in the place of a language
tag, ib. To allow for an alternative datatype URI, the RDF specification would have
to be amended. However, as a sub-datatype of rdf:langString, the constraints of BCP
47 would still apply. It thus seems easier to propose a change to BCP 47: namely to
562
Proceedings of eLex 2019
allow, for the privateuse sub-tag only, the following characters: [-:/a-zA-Z0-9]. This
would then render a language tag of the form "jannaie"@x-deaf:fro/gallo. To be RDF-
compatible, the namespace for x-deaf: would have to be defined in the same RDF
document in which the language tag is used.
We considered creating a user-defined simple XML Schema datatype, as a restriction
on an existing datatype (Carroll & Pan, 2006). Although it would not render a language
tagged string literal, it would render a string literal with an encoded URI: "jannaie"^^
<http://example.org/simpleTypes#froGallo>. However, the URI, although it clearly
identifies the language, would not be dereferenceable which is in opposition to one of
the principles of LD. Furthermore, it would not be appropriate for use when modelling
data using Ontolex-Lemon because the latter requires rdf:langString when representing
forms. This leads us to conclude that Part 4 is required in our pattern, i.e., for the
inclusion of a URI shortcode in the privateuse portion of a language tag, which can
then be mapped to a URI.
Apart from the question of how to design the language tags, a further question arises:
is the granularity of our approach sufficient for the following scenarios? The language
of a linguistic resource, e.g., a text or a dictionary, is written:
1. during a time span or covering a time span, e.g., a collection of 19
th
century
legal documents or a dictionary covering several centuries such as the DEAF,
2. at different times, e.g., the Roman de la Rose that consists of two parts
(ca.1230; ca.1275) by two authors
21
,
3. in different places or covers several places, some parts (in) region A, some
parts (in) Region B.
The scenarios describe multilingual settings that require multilingual labels (a part of
the RDF standard
22
). Scenarios 1 and 2 can be answered with the range of Part 2 of
our pattern. For scenario 3, two questions arise: how to identify (a) the language(s) of
a triple subject (a lexicon, a lexical entry, etc.), and (b) the language(s) of a literal.
Question (a) is answerable with the property dct:language that has multiple values,
such as <http://example.org/language-1> and <http://example.org/language-2>
respectively (cp. E0 with both ISO 639-1 and ISO 639-3 code). Question (b) is
answerable with multiple literals, i.e., duplicated language-tagged literals for the same
subject and predicate, with a custom language tag for each.
21
http://www.deaf-page.de/bibl/bib99r.php#RoselLangl [11-06-2019].
22
https://www.w3.org/community/bpmlod/wiki/Best_practises_-_previous_notes [12-06-
2019].
563
Proceedings of eLex 2019
6. Interface for Language Tag Generation
A user interface and REST API to both generate and decode language tags, currently
in development, is to be demonstrated at eLex 2019. Language tags can be generated
according to our pattern. For the decoding of language tags, the results are available
in JSON, with natural language, RDF/XML and Turtle syntax to follow. Figure 3
shows the user interface. See https://londisizwe.org/language-tags/ for more
information.
Figure 3: User interface for generating and decoding language tags.
7. Conclusions & Future Work
In this paper, we have discussed how to create language tags when modelling linguistic
data as LD for languages for which ISO 639 does not provide language codes. We have
focused on linguistic resources of French that are of interest for diatopic studies, and
we have chosen exemplary data with a diachronic view, including Old-, Middle- and
Modern French dictionaries and a Modern French linguistic atlas. For each exemplar,
we have created a language tag, in line with a proposed pattern. These language tags
564
Proceedings of eLex 2019
identify the language, its historical language stage, a subset of the language (dialect or
patois) in an unambiguous way. Using a URI shortcode, the language tags can be
reduced to a more user-friendly length. This, however, makes them opaque, whereas
the former is more descriptive but can be long. While the use of encoded URIs affects
human-readability, it remains machine-readable nonetheless.
Extension towards MoLA. In collaboration with C. Maria Keet, the authors have been
working on MoLA, a Model for Language Annotation (Gillis-Webber et al., 2019).
MoLA is a lightweight ontology which allows for languoids (a language family,
language, dialect cluster, or lect) to be represented in RDF. Due to its expressiveness,
including MoLA in the modelling of linguistic resources enables comprehensive
language information to be represented. Future work is, thus, to model the languages
identified in these French resources using MoLA.
Other Resources. We conclude the paper returning to linguistic desiderata: Other
linguistic atlases (of the series ALFR, e.g., Lanher et al., 1979–1988 [Lorraine Romane];
Dondaine & Dondaine, 1972–1991 [Franche-Comté]) and dictionaries should be
evaluated for a future conversion to LD. Valuable dictionaries comprise those covering
particular patois and dialects, the comprehensive dictionary of French regionalisms
(Rézeau, 2001), etc. The modelling of lexicologically rich resources of other kinds is a
further task, including a lexicographer’s standard work for historic botany, the Flore
populaire de la France... (Rolland, 1896–1914), and corpora, e.g., the Corpus Historique
du Substandard Français (CHSF, Thun, 2011).
8. Varieties of French
Figures 4 and 5 show the designations of French varieties, the corresponding
Glottocodes and ISO 639-3 codes, respectively. We define the lists of Old French
varieties given by the FEW (von Wartburg (1922–: Beiheft p.63)) and by the DEAF
as authority lists and exclude all regional varieties listed by other resources (e.g.,
Lexilogos) that are not covered by the FEW- or the DEAF list.
Modern French
/
FEW
Old French /
FEW
Old French /
DEAF
Glottolog
(modern)
ISO 639-3
(modern)
français moderne
français moderne stan1290 fra
ancien français ancien français fro *
moyen français moyen français mid1316 frm *
francien
pik. apik. picard pica1241 ** pcd
hain. hennuyer hain1252
art. artésien arto1238
565
Proceedings of eLex 2019
wallon awallon. wallon wall1255 wln
lütt. alütt. liégeois
nam. anam.
flandr. aflandr. français de la
Flandre française
Lille alill. lill1247
champ. achamp. champenois
lothr. alothr. lorrain lorr1242
norm. anorm. normand norm1245 nrf
agn. anglo-normand angl1258 xno *
hbret. haut-breton gall1275
*
Historical language stage. ** 12 sub-languages incl. ‘hain1252’, ‘arto1238’, ‘lill1247’.
Figure 4: List of French varieties, part 1 (terms in French).
Modern French /
FEW
Old French /
FEW
Old French /
DEAF
Glottolog
(modern)
ISO 639-3
(modern)
ang. angevin ange1244
poit. apoit. poitevin poit1240
saint. saintongeais sant1407
tour. tourangeau
orl. orléanais
bourbonn. abourb. bourbonnais bour1246
bourg. abourg. bourguignon bour1247
Lyon ** lyonnais lyon1243 ***
frcomt. afrcomt. franc-comtois fran1262 ***
franco-italien
Nord-Est
Nord
Nord-Ouest
Ouest
Sud-Ouest
centr. Centre
Est
Sud-Est
Terre Sainte
judfr. Judeofrançais zrp *
*
Historical language stage. ** Sub Savoy. *** Sub Francoprovençalic.
Figure 5: List of French varieties, part 2 (terms in French).
566
Proceedings of eLex 2019
9. References
ATILF CNRS & Université de Lorraine (2015). Dictionnaire du Moyen Français,
version 2015 (DMF 2015). Paris. URL http://www.atilf.fr/dmf/. Accessed: 17-
06-2019.
Baldinger, K., Möhren, F. & Städtler, T. (1971–). Dictionnaire étymologique de
l’ancien français (DEAF). Québec / Tübingen / Berlin: Presses de L’Université
Laval / Niemeyer / De Gruyter. DEAFél: https://deaf-server.adw.uni-
heidelberg.de].
Berners-Lee, T. (2006). Linked Data. World Wide Web Consortium. URL
https://www.w3.org/DesignIssues/LinkedData.html. Accessed: 17-06-2019.
Berschin, H., Felixberger, J. & Goebl, H. (2008). Französische Sprachgeschichte.
Hildesheim / Zürich / New York: Olms.
Bizer, C., Heath, T. & Berners-Lee, T. (2009). Linked Data The Story So Far.
International Journal on Semantic Web and Information Systems, 5, pp. 1–22.
Bouda, P. & Cysouw, M. (2012). Treating Dictionaries as a Linked-Data Corpus. In
C. Chiarcos (ed.) Linked Data in Linguistics. Representing and Connecting
Language Data and Language Metadata. Berlin/Heidelberg, Germany: Springer,
pp. 15–23.
Carroll, J. & Pan, J. (2006). XML schema datatypes in RDF and OWL: W3C Working
Group Note 14 March 2006. URL https://www.w3.org/TR/swbp-xsch-
datatypes/. Accessed: 17-06-2019.
Chambon, J. P. (1997). Pour la localisation d’un texte de moyen français: le Mystère
de Saint Sébastien. In G. Kleiber & M. Riebel (eds.) Les formes du sens: Etudes
de linguistique française, médiévale et générale offertes à Robert Martin à
l’occasion de ses 60 ans. Louvain-la-Neuve: Duculot, pp. 201–216.
Chauveau, J. P. (2016). Régionalismes médiévaux et dialectismes contemporains en
hauteBretagne. In M. Glessgen & D. Trotter (eds.) La régionalité lexicale du
français au Moyen Âge. Strasbourg: ÉLiPhi, pp. 131–166.
Chiarcos, C., McCrae, J., Cimiano, P. & Fellbaum, C. (2013). Towards Open Data for
Linguistics: Lexical Linked Data. In A. Oltramari, P. Vossen & L. Qin et al. (eds.)
New Trends of Research in Ontologies and Lexical Resources: Ideas, Projects,
Systems. Berlin / Heidelberg: Springer, pp. 7–25.
Cyganiak, R., Wood, D. & Lanthaler, M. (2014). RDF 1.1. concepts and abstract
syntax: W3C recommendation 25 February 2014. URL
https://www.w3.org/TR/2014/ REC-rdf11-concepts-20140225/. Accessed: 17-06-
2019. de Melo, G. (2015). Lexvo.org: Language-Related Information for the
Linguistic Linked Data Cloud. Semantic Web, 6(4), pp. 393–400.
Denizot, J. (1910). Vocabulaire patois de Sainte-Sabine et ses environs (Côte-d’Or).
Beaune: Imprimerie Beaunoise.
Dondaine, C. & Dondaine, L. (1972–1991). Atlas linguistique et ethnographique de la
Franche-Comté (ALFC). Paris: Éd. du CNRS.
567
Proceedings of eLex 2019
Gally, S., Chauvin-Payan, C. & Davoine P. A. et al. (2013). GéoDialect : Exploration
des outils géomatiques pour le traitement et l’analyse des données
géolinguistiques. Géolinguistique, 14, pp. 186–208.
Gillis-Webber, F. & Tittel, S. (2019). The Shortcomings of Language Tags for Linked
Data when Modeling Lesser-Known Languages. In Proceedings of LDK2019,
Leipzig, Germany, 21-22 May 2019, OASIcs, Vol. 70. pp. 4:1–4:15.
Gillis-Webber, F., Tittel, S. & Keet, M. (2019). A Model for Language Annotations on
the Web. In B. Villazón-Terrazas & Y. Hidalgo-Delgado (eds.) Knowledge Graphs
and Semantic Web. 1st Iberoamerican Conference, KGSWC 2019, Villa Clara,
Cuba, June 23-30, 2019, Proceedings. pp. 1–16.
Gilliéron, J. & Edmont, E. (1902–1910). Atlas linguistique de la France. Paris:
Champion. Glessgen, M. & Trotter, D. (2016). La régionalité lexicale du français
au Moyen Âge. Strasbourg: ÉLiPhi.
Gleßgen, M. D. & Thibaut, A. (2005). La «régionalité linguistique»: essai définitoire.
In M.D. Gleßgen & A. Thibaut (eds.) La lexicographie différentielle du français
et le Dictionnaire des régionalismes de France. Presses Univ. de Strasbourg, pp.
III–XVII.
International Organization for Standardization (n.d.). Language codes – ISO 639. URL
https://www.iso.org/iso-639-language-codes.html. Accessed: 17-02-2019.
International Phonetic Association (2005). International Phonetic Alphabet. Tech. rep.
URL https://www.internationalphoneticassociation.org/. Accessed: 17-02-2019.
Lanher, J., Litaize, A. & Richard, J. (1979–1988). Atlas linguistique et ethnographique
de la Lorraine Romane (ALLR). Paris: Éd. du CNRS.
Millot, C. (1905–1922 (edition 1998)). Dictionnaire de patois de Mancey. Tournus:
Société des amis des arts et des sciences de Tournus.
Moran, S. (2012). Using Linked Data to Create a Typological Knowledge Base. In C.
Chiarcos (ed.) Linked Data in Linguistics. Representing and Connecting
Language Data and Language Metadata. Springer, pp. 129–138.
Möhren, F. (2016). La régionalité dans le DEAF historique et programme. In M.
Glessgen & D. Trotter (eds.) La régionalité lexicale du français au Moyen Âge.
Strasbourg: ÉLiPhi, pp. 37–50.
Phillips, A. & Davis, M. (2009). Tags for Identifiying Languages. BCP, 47. URL
https://tools.ietf.org/html/bcp47. Accessed: 17-06-2019.
Renders, P. (2016). La régionalité lexicale du moyen français (1350–1500). In M.
Glessgen & D. Trotter (eds.) La régionalité lexicale du français au Moyen Âge.
Strasbourg: ÉLiPhi, pp. 85–96.
Rickard, P. (1974). A history of the French language. London: Hutchinson University
Library.
Rolland, E. (1896–1914). Flore populaire de la France ou histoire naturelle des plantes
dans leurs rapports avec la linguistique et le folklore. Paris: Rolland.
Rézeau, P. (ed.) (2001). Dictionnaire des régionalismes de France. Géographie et
histoire d’un patrimoine linguistique. Bruxelles: De Boeck.
568
Proceedings of eLex 2019
Rézeau, P. (2007). Richesses du français et géographie linguistique. Bruxelles: De Boeck
& Larcier.
Rézeau, P. (2016). La régionalité lexicale du français après 1500, à travers des
régionalismes recueillis dans les correspondances de poilus. In M. Glessgen & D.
Trotter (eds.) La régionalité lexicale du français au Moyen Âge. Strasbourg:
ÉLiPhi, pp. 111–130.
Séguy, J. (1973). Les Atlas linguistiques de la France par régions. Langue Française,
18, pp. 65–90.
Thun, H. (2011). Die diachrone Erforschung der français régionaux auf der Grundlage
des Corpus Historique du Substandard Français. In C. Schlaak & L. Busse (eds.)
Sprachkontakte, Sprachvariation und Sprachwandel. Narr, pp. 359–394.
Tittel, S. (2016). La régionalité lexicale de l’ancien français (ca.1100 – ca.1350) : Une
enquête sur la base du Dictionnaire étymologique de l’ancien français. In M.
Glessgen & D. Trotter (eds.) La régionalité lexicale du français au Moyen Âge.
Strasbourg: ÉLiPhi, pp. 61–84.
Tittel, S., Bermúdez-Sabel, H. & Chiarcos, C. (2018). Using RDFa to Link Text and
Dictionary Data for Medieval French. In J. P. McCrae, C. Chiarcos & T. Declerck
et al. (eds.) Proceedings of the Eleventh International Conference on Language
Resources and Evaluation (LREC 2018). 6th Workshop on Linked Data in
Linguistics (LDL-2018), 12 May 2018, Miyazaki, Japan. Paris: ELRA, pp. 30–38.
Tittel, S. & Chiarcos, C. (2018). 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. Paris:
ELRA, pp. 58–66.
Vannini, L. & Le Crosnier, H. (2012). Net.lang. Towards the multilingual cyberspace.
Caen: C & F Éditions.
Varlet, M. (1896). Dictionnaire du patois meusien. Verdun: Société Philomathique de
Verdun.
Vasseur, G. (1998). Dictionnaire des parlers picards du Vimeu (Somme), avec index
français-picard. Fontenay-sous-Bois: SIDES.
von Wartburg, W. (1922–). Französisches Etymologisches Wörterbuch (FEW). Bonn,
Heidelberg, Leipzig/Berlin, Basel: ATILF. [Continued by O. Jänicke, C. T.
Gossen, J. P. Chambon, J.-P. Chauveau, and Yan Greub].
Wolf, H.J. (1979). Französische Sprachgeschichte. Heidelberg: Quelle u. Meyer.
This work is licensed under the Creative Commons Attribution ShareAlike 4.0
International License.
http://creativecommons.org/licenses/by-sa/4.0/
569
Proceedings of eLex 2019
... Nev- ertheless, NIF has been used as a publication format for corpora with entity annotations. 121 NIF continues to be a popular component of the DBpedia technology stack. At the same time, active development of NIF seems to have slowed down since the mid-2010s, whereas limited progress on NIF standardization has been achieved. ...
... The core data structure of the Web Annotation Data Model is the annotation, i.e., instances of oa:Annotation that have an oa:hasTarget property that identifies the element that carries the annotation, and the oa:has- 121 The most prominent example, the NIF edition of the Brown corpus published in 2015, formerly available from http://brown. nlp2rdf.org/, ...
... lang-subtags-templates.xhtml 166 Cf. https://github.com/w3c/i18n-discuss/issues/13 . 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). ...
Article
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.
... Searching for and modelling diachronic change requires rethinking some contemporary (Semantic) Web infrastructure. As [190] shows, standardised language tags cannot capture the differences between Old-, Middle-and Modern French resources. Digital editions, often modelled in TEI [191], are a rich resource of diachronic language variation. ...
Article
Full-text available
This paper presents an overview of the LL(O)D and NLP methods, tools and data for detecting and representing semantic change, with its main application in humanities research. The paper’s aim is to provide the starting point for the construction of a workflow and set of multilingual diachronic ontologies within the humanities use case of the COST Action Nexus Linguarum, European network for Web-centred linguistic data science, CA18209. The survey focuses on the essential aspects needed to understand the current trends and to build applications in this area of study.
Conference Paper
Full-text available
The identification and annotation of languages in an unambiguous and standardized way is essential for the description of linguistic data. It is the prerequisite for machine-based interpretation, aggregation, and re-use of the data with respect to different languages. This makes it a key aspect especially for Linked Data and the multilingual Semantic Web. The standard for language tags is defined by IETF's BCP 47 and ISO 639 provides the language codes that are the tags' main constituents. However, for the identification of lesser-known languages, endangered languages, regional varieties or historical stages of a language, the ISO 639 codes are insufficient. Also, the optional language sub-tags compliant with BCP 47 do not offer a possibility fine-grained enough to represent linguistic variation. We propose a versatile pattern that extends the BCP 47 sub-tag privateuse and is, thus, able to overcome the limits of BCP 47 and ISO 639. Sufficient coverage of the pattern is demonstrated with the use case of linguistic Linked Data of the endangered Gascon language. We show how to use a URI shortcode for the extended sub-tag, making the length compliant with BCP 47. We achieve this with a web application and API developed to encode and decode the language tag.
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.
Article
Full-text available
La variation diatopique du lexique français médiéval constitue une dimension linguistique à part entière qui est depuis toujours largement sous-estimée – et sous-exploitée – par les historiens de la langue, les philologues, les éditeurs de texte et même les lexicographes. Le présent volume souhaite dépasser les approches centenaires qui ont considéré la régionalité lexicale essentiellement au niveau microscopique des lexèmes ou des textes individuels. Son objectif est d’appréhender le phénomène de manière systématique en le plaçant dans le cadre interprétatif de la linguistique variationnelle actuelle. Le volume comporte deux volets: (i) un inventaire de 2 800 lexèmes régionaux provenant du territoire d’oïl continental et établi notamment à partir de l’œuvre intégrale de Gilles Roques, en l’honneur duquel le colloque a été réalisé; (ii) une série d’études thématiques qui tentent d’identifier la place de la régionalité lexicale dans la lexicographie du français, dans l’évolution du diasystème entre le 9e et le 20e siècle et dans les différents genres textuels (littérature profane et religieuse, textes documentaires). Ils mettent également en relief le rôle du contact linguistique pour la régionalité lexicale et le rôle de cette dernière dans la formation d’une variété standard en comparant le français avec les langues voisines (anglais, espagnol et italien).
Article
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.
Article
Full-text available
Lexvo.org 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.
Article
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.