Incorrect prefLabel chosen when multiple language labels are defined

[mdorf]

When ontology defines RDFS labels in multiple languages for a given class, BioPortal picks a random (the first one in the array) label to be designated as the prefLabel.

Example: TXPO ontology (this ontology is private, must be an admin to view). Some of the classes are displayed in English and some are in Japanese.

[mdorf]

This ontology does not explicitly set the property: http://www.w3.org/2004/02/skos/core#prefLabel, which means, the prefLabel property is determined using the http://www.w3.org/2000/01/rdf-schema#label property. The latter may contain multiple entries, one for each specified language. Ex:

<rdfs:label xml:lang="en">occurrent</rdfs:label>
<rdfs:label xml:lang="ja">生起物</rdfs:label>

The code that sets the prefLabel property is located here:

https://github.com/ncbo/ontologies_linked_data/blob/master/lib/ontologies_linked_data/models/ontology_submission.rb#L648

The code that simply picks the first label on the list is located here:

https://github.com/ncbo/ontologies_linked_data/blob/master/lib/ontologies_linked_data/models/ontology_submission.rb#L652

The code on line 652 must instead examine whether an instance of :EN language label exists and pick its value instead of picking the first value in the array.

[mdorf]

This issue is related to the previously addressed issue in JIRA: https://jira.bmir.stanford.edu/browse/NCBO-1662

[jonquet]

We have fixed this in SIFR and AgroPortal. We have a parameter to choose the default language.
@vemonet might certainly point to the change in code.

[graybeal]

If I understand our status correctly we have it working (to just show English first) for skos:prefLabel, but not for rdfs:label.

@vemonet It would be great to have a pointer to your code for the choice of default language. I assume that works for both rdfs:label and skos:prefLabel?

Also, I see the issue in annotations when annotating the default string, the 10th or so row for 'melanoma' puts up a Chinese label.

[graybeal]

@jonquet are you in a position that you could make a pull request to make it possible to choose the default language?

[jvendetti]

An end user reported another instance of this issue today for the ontology with acronym TERRADCAT_AP. One example is the class with an English language label of "Distribution". BioPortal displays the Arabic language label in the class tree:

This particular class comes from the imported DCAT ontology. DCAT is maintained in GitHub where you can view the declaration for the Distribution class, along with the multiple language labels:

dcat:Distribution
  a rdfs:Class ;
  a owl:Class ;
  ...
  rdfs:label "Distribuce"@cs ;
  rdfs:label "Distribución"@es ;
  rdfs:label "Distribution"@en ;
  rdfs:label "Distribution"@fr ;
  rdfs:label "Distribuzione"@it ;
  rdfs:label "Διανομή"@el ;
  rdfs:label "التوزيع"@ar ;
  rdfs:label "配信"@ja ;
  rdfs:label "Distribution"@da ;
  ...

[jonquet]

@jonquet are you in a position that you could make a pull request to make it possible to choose the default language?

No. This is 7 years old now.

[vemonet]

Hi @jonquet and @graybeal , sorry never noticed the notifications from this issue before

I remember we had some Java to clean ontologies with multiple labels, but not sure about choosing the default language, would need more details/example

In this case it seems to be a problem specifically related to rdfs:label not being used as a class label like skos:prefLabel.

You can fix it via the calls done to resolve labels when skos:prefLabel is not used if the lang attribute is accessible: https://github.com/ncbo/ontologies_linked_data/blob/master/lib/ontologies_linked_data/models/ontology_submission.rb#L645

But it would probably be better to fix it downstream, when the main prefLabel is picked, so you don't need to rely on this call everytime the class is accessed

[jonquet]

We are working on anew strategy/algorithm for this and will keep you updated soon.