June 30, 2023 Project deliverable Open Access

Interlingual Index for the EASIER Project's Core Sign Languages

Bigeard, Sam; Schulder, Marc; Kopf, Maria; Hanke, Thomas; Vasilaki, Kiriaki; Vacalopoulou, Anna; Goulas, Theodor; Dimou, Athanasia-Lida; Fotinea, Stavroula-Evita; Efthimiou, Eleni; Fox, Neil; Crasborn, Onno; Westenberg, Lianne; Ebling, Sarah; Wawrinka, Laure

The purpose of the inter-lingual index is to link the lexical resources from the different languages
of the project and make them machine-readable. The earlier deliverable D6.3 was the first
version of this index. It included German Sign Language (DGS) and Greek Sign Language
(GSL). This deliverable is the second version of the index. It covers further core sign languages
of the project: British Sign Language (BSL), Sign Language of the Netherlands (NGT), French
Sign Language (LSF) and Swiss-German Sign Language (DSGS). The next version will be
deliverable 6.5 and will include languages beyond the project’s core languages.
The deliverable is the index itself. This report provides background information on wordnet
research, explains our method and choices, and presents the resulting dataset.
Our interlingual index uses the wordnet concept of synonym sets (synsets), which define con-
cepts by gathering signs and words that can represent the same meaning. This approach is
more resistant to translation mistakes stemming from translation pairs being only valid for cer-
tain word/sign meanings. It also provides a new way to define sign types that does not rely on
approximate translations to a single spoken language word, the way glosses do. As a basis for
our index, we build on the synset inventory of Open Multilingual Wordnet (OMW).
We use a three-step method: The first step is automatically matching candidate synsets to signs
using the keywords and glosses associated with the sign. The second step is automatically
validating links that are most likely to be correct. The final step is manual validation of the
remaining links, prioritising the most useful signs.
This work has resulted in a dataset of 7929 signs in 6 sign languages linked to 11806 synsets.
Additionally, a web interface has been launched to make the index accessible for the general
public.