A First Corpus of AZee Discourse Expressions

Proceedings of the 13th Conference on Language Resources and Evaluation (LREC 2022), pages 1560–1565 Marseille, 20-25 June 2022 © European Language Resources Association (ELRA), licensed under CC-BY-NC-4.0 1560 A First Corpus of AZee Discourse Expressions Camille Challant, Michael Filhol Universit´e Paris-Saclay, CNRS, Laboratoire Interdisciplinaire des Sciences du Num´erique, 91400, Orsay, France camille.challant@lisn.upsaclay.fr, michael.filhol@cnrs.fr Abstract This paper presents a corpus of AZee discourse expressions, i.e. expressions which formally describe Sign Language utterances of any length using the AZee approach and language. The construction of this corpus had two main goals: a first reference corpus for AZee, and a test of its coverage on a significant sample of real-life utterances. We worked on productions from an existing corpus, namely the 40 br`eves, containing an hour of French Sign Language. We wrote the corresponding AZee discourse expressions for the entire video content, i.e. expressions capturing the forms produced by the signers and their associated meaning by combining known production rules, a basic building block for these expressions. These are made available as a version 2 extension of the 40 br`eves. We explain the way in which these expressions can be built, present the resulting corpus and set of production rules used, and perform first measurements on it. We also propose an evaluation of our corpus: for one hour of discourse, AZee allows to describe 94% of it, while ongoing studies are increasing this coverage. This corpus offers a lot of future prospects, for instance concerning synthesis with virtual signers, machine translation or formal grammars for Sign Language. Keywords: AZee, Sign Language, Formal representation, Corpus 1. Introduction Formal Sign Language (SL) description is often an essen- tial piece for any kind of processing, especially for synthe- sis with virtual signers. The most common way to repre- sent SL is based on models developed for spoken languages which have a writing system. They are concerned with the order of constituents in the discourse, thus assuming the ex- istence of different linguistic levels (phonological, lexical, syntactic, etc.) and a sequence order (Napoli and Sutton- Spence, 2014; Padden, 2016; Pfau et al., 2018). This raises several problems, for instance because some structures in SL are difficult to identify as clearly belonging to a specific linguistic level, or because the multi-linearity of SL and a linear sequence are not always compatible. This paper deals with work done on a formal model named AZee (Filhol et al., 2014), which avoids most assumptions on language if only coming from spoken languages. The basic principle assumed is that a language creates strong associations be- tween forms and meaning. It allows to take into account many specific features of SL, related to their visual-gestural modality. As AZee began to prove itself for synthesis with an avatar on short but increasingly complex examples (Fil- hol and McDonald, 2020), we wanted to scale up the test and address SL discourse longer in duration. The first goal of our study is therefore to test the coverage of AZee, by confronting the model with a large …

Content truncated to 500 words