New pilot dataset: Ghanaian Sign Language as Linked Open Data (OntoLex-Lemon + sldc ontology)

[Eunice Esi Essuman]

Dear Colleagues,

I wanted to share a small pilot project I have been working on at the intersection of sign language linguistics and Linguistic Linked Open Data.

I have built what I believe is the first RDF representation of Ghanaian Sign Language (GSL) vocabulary, modelling 18 signs from the Fragkiadakis, Nyst & Nyarko (2021) GSL Lexicon (Zenodo, CC-BY 4.0) using OntoLex-Lemon and the Declerck (2022) sign language ontology (sldc). The dataset includes 353 triples and a non-manual marker extension (gsl: NonManualMarker) covering facial expression, mouth pattern, and head movement — features that the SLDc ontology flags as an open modelling problem.

The most interesting result: a single SPARQL query retrieves WHERE and WHAT as a grammatical class — not because they are labelled as WH-question signs, but because they share a non-manual marker (raised eyebrows) encoded independently on each sign. This is impossible in a gloss-only dataset.

The dataset is openly published:
— GitHub: https://github.com/LINGUISTEUNICE/gsl-linked-data

— Zenodo DOI: https://zenodo.org/records/20961293 

This is a pilot, not a complete solution. I built this to understand the problem from the inside. I am aware of the limitations (string literals for location values, no linkage to pose data, small scale) and have documented them honestly in the README.

I would welcome any feedback from the community, particularly on the NonManualMarker modelling approach and whether there are existing LLOD resources for sign languages I should be linking to.

With thanks,
Eunice