clld app serving the data from clld/concepticon-data
JavaScript Python Mako CSS
Clone or download

Concepticon in CLLD

Since the data and mappings provided by the concepticon project are an important building block to interoperability between lexical resources, they should be published as clld app (edited by Cysouw and List), thereby providing an API which can be used by tools and methods.

Data model

The main entities of the data model are

  • ConcepticonConcept: An "authoritative" or "hub" concept, which can be used to relate other concepts to or directly by referencing it from wordlists. A ConcepticonConcept is identified by its unique English gloss. All ConcepticonConcepts have a description or example, possibly imported from OmegaWiki. (maps to Parameter)
  • ConceptList: A list of concepts typically created to collect wordlists; e.g. the swadesh list, but maybe also OmegaWiki's DefinedMeanings. (maps to Contribution)
  • SourceConcept: A concept as it appears in a ConceptList. (maps to ValueSet, the relation from ValueSet to Parameter is interpreted as the canonical sameAs relation (see below).)
  • Relation (better term?): Relates SourceConcepts to ConcepticonConcepts, specifying a relation from "sameAs|narrower|broader". Each SourceConcept is assumed to be related to exactly one ConcepticonConcept via a sameAs relation.

clld app

With this data model, a clld app can provide:

  • URLs identifying SourceConcepts and ConcepticonConcepts (as well as ConceptLists).
  • an API allowing discovery of related concepts given either a Source- or ConcepticonConcept.

Data curation

Data curation could proceed pretty much as done now, i.e. the data can be curated as a set of csv files maintained in a GitHub repository. The clld app can import well-defined releases of this repository. The set of files may look as follows:

  • One csv file listing ConcepticonConcepts: ID Gloss Description
  • One csv file per ConceptList, e.g. swadesh-1955.csv: ID Gloss ConcepticonConceptID
  • One csv file per ConceptList specifying additional relations: ID ConcepticonConceptID relation

The data repository could be equipped with a small python package providing data integrity checks which can be run after each push as Travis-CI jobs.