AGROVOC

From Ontology-Lexica Community Group

AGROVOC

requirements covered:: most of the vocabulary modules we are developing for OntoLex

refer to: Armando Stellato

AGROVOC is a controlled vocabulary covering all areas of interest to FAO, including food, nutrition, agriculture, fisheries, forestry, environment etc. To date, AGROVOC contains over 32,000 concepts organized in a hierarchy, each concept may have labels in up to 25 languages: Arabic, Chinese, Czech, English, French, German, Hindi, Hungarian, Italian, Japanese, Korean, Lao, Persian, Polish, Portuguese, Russian, Slovak, Spanish, Thai, Turkish, Malaysian, Moldavian, Telugu, Ukrainian.

Current metadata info about agrovoc are published here: http://aims.fao.org/aos/agrovoc/void.ttl

Being part of the group maintaining AGROVOC, Armando Stellato can introduce the OntoLex model to the FAO group maintaining AGROVOC and, if OntoLex goes into a promising direction, we could early adopt it for modeling the linguistic aspects of the thesaurus.

The property vocabulary of VOCBENCH is http://aims.fao.org/aos/agrontology. There is wide margin of improvement in this vocabulary, as this has been only recently separated from the thesaurus data (previously, data and vocabulary were put together under the common umbrella of AGROVOC) and also still includes many modeling choices taken in the past, before even SKOS and SKOS-XL were developed (and before many popular ontologies were distributed). As you can see by inspecting Agrontology, there are several lexical relations mapping lexical entries between them. Lexical Entries are currently modeled as SKOS-XL Labels, thus first class citizens (for which at least editorial data is always present, other than domain attributes and relationships)

List of properties for which a better and clearer definition would be welcome

Here follows a list of properties adopted in Agrovoc which could be replaced by the ontolex vocabulary. In some cases, this may require simply a 1-1 replacement (the benefit would be at least to adopt a standard vocabulary), while in others, some structural changes may be necessary. A complete description of the properties can be read on: http://aims.fao.org/aos/agrontology

For those willing to look at the data, a sparql end point is available at: http://202.45.142.113:10035/repositories/agrovoc


Object Properties between reified labels (subproperties of skosxl:labelRelation):

correspondsTo                 | seems empty, and also wrongly modeled, proposed for removal
hasAbbreviation               | seems interesting for OntoLex
hasAcronym                    | it's a subproperty of hasAbbreviation. Since hasAbbreviation is too a concrete property, I'm checking differences between them
hasBroaderSynonym             
hasChemicalFormula            | should be attached to a concept, not to a label    
hasDialectalVariant           
hasLocalName                  
hasNarrowerSynonym            
hasNearSynonym                
hasOldName                    
hasRelatedTerm                
hasScientificName              
hasSymbol                     
hasSynonym                    
hasTradeName/isTradeNameOf    | asking2FAO : btw, seems to be a wrong use of a property, as it should involve more complex constructs
hasTranslation                
hasTransliteration            
isAbbreviationOf              
isAcronymOf                   
isChemicalFormulaOf           | see hasChemicalFormula
isDialectalVariantOf          
isLocalNameOf                 
isOldNameOf                   
isSymbolFor                                  
isTransliterationOf                             
scientificNameOf              * see below          

Regarding scientificNameOf: this should be actually a property relating to concepts. Apart from that, I read many posts about suggestions for scientificName. Though quite specific (not sure it is covered by the objectives of our model, as we maybe interested more in provide top properties for linguistic and ontolinguistic modeling), it maybe worth to define a way for covering scientific names, which are very common in thesauri related to nature.

For those willing to explore these properties, I wrote a simple query which can be run on the previously mentioned SPARQL endpoint:

select ?l1 ?l2 (lang(?l1) as ?language) {
  ?a <http://aims.fao.org/aos/agrontology#hasTradeName> ?b .
  ?a <http://www.w3.org/2008/05/skos-xl#literalForm> ?l1 .
  ?b <http://www.w3.org/2008/05/skos-xl#literalForm> ?l2 .
  FILTER (lang(?l1) = "en" || lang(?l1) = "it")
}


Datatype Properties of reified labels:

hasTermType                        
hasPlural                          
hasSingular                        
hasSpellingVariant                 
hasStemmedForm                     
hasTermVariant