This document is aimed to guide in the process of creating a linked data (LD) version of a lexical resource, particularly a bilingual dictionary. It contains advice on the vocabulary selection, RDF generation process, and publication of the results. As a result of publishing the data as LD, the converted language resource will be more interoperable and easily accessible on the Web of Data by means of standard Semantic Web technologies. The process described in this document has been illustrated with real examples extracted from Apertium RDF, an open-source machine translation system which has their data available for download.

There are a number of ways that one may participate in the development of this report:

• Mailing list: public-bpmlod@w3.org
• Wiki: Main page
• More information about meetings of the BPMLOD group can be obtained here
• Source code for this document can be found on Github.

Description of the type of resource

The type of language resources covered in this document is bilingual electronic dictionaries. A bilingual dictionary provides translation equivalents of words and phrases (and examples of usage) of one language in another language. They can be unidirectional or bidirectional, allowing translation, in the latter case, to and from both languages. In addition to the translation, a bilingual dictionary usually indicates the part of speech, as well as the gender, verb type, declination model and other semantic, syntactic, and grammatical properties to help a non-native speaker use the word.

We refer to bilingual dictionaries that have data in a machine-processable format, no matter whether it is stored locally or is accessible on the Web (e.g. for download). We assume that the data is represented in a structured or semi-structured way (e.g. relational database, XML, CSV, etc.).

The topic is illustrated by examples from the conversion of the Apertium dictionaries into RDF [AP_RDF]. Apertium [AP_PAPER] is a free/open-source machine translation platform originally designed to translate between closely related languages, although it was later expanded to treat more divergent language pairs. The Lexical Markup Framework [LMF] versions of their linguistic data are available here and they were used as starting point for the first RDF version of the Apertium dictionaries [AP-RDF_PAPER], based on the original lemon model. A more recent version of Apertium RDF was later developed by following Ontolex lemon [AP_RDF_V2_PAPER].

Selection of vocabularies

• We propose Ontolex lemon (LExicon Model for ONtologies) [ONTOLEX, ONTOLEX_PAPER] to model the RDF representation of the linguistic descriptions contained in bilingual dictionaries. Ontolex lemon has been designed to extend the lexical layer of ontologies with as much linguistic information as needed, and to provide it as LD on the Web. It is made up of a series of modules, of which we use the following three:
  • core, where lexical entries, forms, and lexical senses are defined,
  • lime, where lexicon is defined,
  • vartrans. We mainly use two classes: Translation and TranslationSet. Translation is a reification of the relation between two lexical senses associated to entries in different languages. Using a reified class allows us to describe attributes of the Translation object, such as source and target of the translation, category, to determine the type of translaiton, as well as attributes denoting provenance (see below). TranslationSet is a group of translations, typically those coming of the same bilingual dictionary.
• The use of lemon is complemented with Lexinfo [LEXINFO]. Lexinfo is an ontology of types, values and properties that is used with the lemon model, partially derived from ISOcat. We use Lexinfo as a catalog of data categories (e.g. to denote gender, number, part of speech, etc.).
• Translation categories are represented by pointing to an external catalog (e.g. to state that a translation is a "cultural equivalent"). We propose the one at [TRCAT] but any other could be used instead.
• Other extendedly used vocabularies such as Dublin Core [DC] are used to attach valuable information about provenance, authoring, versioning, or licensing.
• Finally, the Data Catalogue Vocabulary [DCAT] is used to represent other metadata information associated to the publication of the RDF dataset.

RDF generation

For the generation and publication processes we have followed existing recommendations [GUIDE_MLD, CONV], adapted to this particular case.

	
<Lexicon>
   <feat att="language" val="en"/>
   ...
   <LexicalEntry id="bench-n-en">
      <feat att="partOfSpeech" val="n"/>
	  <Lemma>
	     <feat att="writtenForm" val="bench"/>
	  </Lemma>
      <Sense id="bench_banco-n-l"/>
   </LexicalEntry>
   ...
</Lexicon>
<Lexicon>
    <feat att="language" val="es"/>
    ...
    <LexicalEntry id="banco-n-es">	
	   <feat att="partOfSpeech" val="n"/>
	   <Lemma>
	       <feat att="writtenForm" val="banco"/>
	   </Lemma>
	   <Sense id="banco_bench-n-r"/>
    </LexicalEntry>
    ...
</Lexicon>
...
<SenseAxis id="bench_banco-n-banco_bench-n" senses="bench_banco-n-l banco_bench-n-r"/>
...
	

Modelling

The first step in the modelling phase is the selection of the domain vocabularies to be used, as described in Section 2 (Selection of vocabularies). The next thing is to decide how the representation scheme of the source data has to be mapped into the new model. In the case of bilingual dictionaries, each dictionary is converted into three different objects in RDF (regardless if the original data comes in one or several files):

• Source lexicon
• Target lexicon
• Translation Set

This is illustrated in Figure 1, for the conversion of an English-Spanish bilingual dictionary.

In our view, this division fits more naturally in the scheme of ontolex and the vartans module. As a result, two independent monolingual lexicons will be published on the Web of Data, along with a set of translations that connects them. The publication of additional bilingual dictionaries (following the same scheme) would imply the creation of a pool of online monolingual lexicons that grows with time, all of them potentially connected within the same RDF graph by sets of translations.

Going into the details of the model, Figure 2 illustrates the representation scheme used for a single translation, in terms of ontolex and the vartans module:

In short, ontolex:LexicalEntry and its associated properties are used to account for the lexical information, while the vartrans:Translation class puts them in connection through ontolex:LexicalSense. Other simpler options are also possible, such as connecting the lexical entries directly without defining "intermediate" senses through the vartrans:translatableAs property (see VT). Nevertheless, translations occur between specific meanings of the words and the use of the vartrans:Translation as a connector between two ontolex:LexicalSense allows us to represent this explicitly.

	
English lexicon: http://linguistic.linkeddata.es/id/apertium/lexiconEN 
Spanish lexicon: http://linguistic.linkeddata.es/id/apertium/lexiconES 
English-Spanish translation set: http://linguistic.linkeddata.es/id/apertium/tranSetEN-ES 

	

apertium:lexiconEN a lime:Lexicon ;
	dc:source <http://hdl.handle.net/10230/17110> .
...
apertium:lexiconEN lime:entry apertium:lexiconEN/bench-n-en .
apertium:lexiconEN/bench-n-en a ontolex:LexicalEntry ;
	ontolex:lexicalForm apertium:lexiconEN/bench-n-en-form ;
	lexinfo:partOfSpeech lexinfo:noun .
apertium:lexiconEN/bench-n-en-form a ontolex:Form ;
	ontolex:writtenRep "bench"@en .
	

apertium:lexiconES a lime:Lexicon ;
	dc:source <http://hdl.handle.net/10230/17110> .
...
apertium:lexiconES lime:entry apertium:lexiconES/banco-n-es .
apertium:lexiconES/banco-n-es a ontolex:LexicalEntry ;
	ontolex:lexicalForm apertium:lexiconES/banco-n-es-form ;
	lexinfo:partOfSpeech lexinfo:noun .
apertium:lexiconES/banco-n-es-form a ontolex:Form ;
	ontolex:writtenRep "banco"@es .
	
	

apertium:tranSetEN-ES a vartrans:TranslationSet ;
	dc:source <http://hdl.handle.net/10230/17110> ;
...	
apertium:tranSetEN-ES vartrans:trans apertium:tranSetEN-ES/bench_banco-n-en-sense-banco_bench-n-es-sense-trans .
apertium:tranSetEN-ES/bench_banco-n-en-sense a ontolex:LexicalSense ;
	ontolex:isSenseOf apertium:lexiconEN/bench-n-en .
apertium:tranSetEN-ES/banco_bench-n-es-sense a ontolex:LexicalSense ;
	ontolex:isSenseOf apertium:lexiconES/banco-n-es .
apertium:tranSetEN-ES/bench_banco-n-en-sense-banco_bench-n-es-sense-trans a vartrans:Translation ;
	vartrans:source apertium:tranSetEN-ES/bench_banco-n-en-sense ;
	vartrans:target apertium:tranSetEN-ES/banco_bench-n-es-sense .

	

Publication

The publication step involves: (1) dataset publication, (2) metadata publication, and (3) enabling effective discovery. Here we focus on the second task (metadata publication). In the context of LD, there are two major vocabularies for metadata of datasets and catalogs: VoID (Vocabulary of Interlinked Datasets) [VOID], and DCAT (Data Catalogue Vocabulary) [DCAT]. In our running example, we use DCAT for describing the elements generated in the RDF conversion of bilingual dictionaries. In any case, DCAT can be complemented with VoID or other vocabularies if required.

--------------- TO BE UPDATED... ---------------

The RDF version of Apertium EN-ES was published in Datahub. The Datahub platform created a metadata file for the Apertium EN-ES dataset based on DCAT. We extended such metadata file with additional missing information such as provenance, license, and related resources. The extended metadata was published as part of the Apertium EN-ES Datahub entry. The following lines are a fragment of it:

<dcat:Dataset rdf:about="http://linguistic.linkeddata.es/set/apertium/EN-ES">
   <owl:sameAs rdf:resource="http://datahub.io/dataset/apertium-en-es"></owl:sameAs>
   <dct:source rdf:resource="http://hdl.handle.net/10230/17110"></dct:source>
   <dct:license rdf:resource="http://purl.oclc.org/NET/rdflicense/gpl-3.0"></dct:license>
   <rdfs:seeAlso rdf:resource="http://dbpedia.org/resource/Apertium"></rdfs:seeAlso>
   <rdfs:seeAlso rdf:resource="http://purl.org/ms-lod/UPF-MetadataRecords.ttl#Apertium-en-es_resource-5v2"></rdfs:seeAlso>
</dcat:Dataset>
	

Recommendations

• Separate the monolingual lexicons from the translation sets (different graphs and/or files).
• Lexical senses should play the role of connectors between translations and lexical entries.
• Be consistent with the rules for naming and URIs creation.
• Keep the identifiers of the legacy data if possible, but remove indicators of directionality if any (e.g. "l", "r", "left", "right", ...)