This paper investigates the requirements for intelligible high-quality motion capture of sign language, with a particular focus on face animation. There are several methods for capturing facial animation, yet there is no established standard for which type of facial animation is the most optimal for sign languages. We compare two facial animation approaches — ARKit and Metahuman Animator (MHA) in terms of intelligibility. As expected, a human evaluation study with deaf Swedish Sign Language signers showed that MHA outperforms ARKit because MHA has a higher number of facial controls and an advanced depth data solver. In addition, we introduce a biLSTM-based occlusion infilling technique for MHA data as opposed to linear interpolation. Although biLSTM infilled MHA animations showed no improvement in human evaluation compared to original MHA animations, the model produces linguistically reasonable infillings upon qualitative exploration of the renders.

Stretching over the northernmost parts of Europe, with Norway, Finland, and Denmark as its closest neighbors, Sweden remains a sparsely populated (10.5 million inhabitants) and linguistically a fairly uniform country. It is the cradle of the world famous social-democratic “welfare state,” which emerged over the course of the 20th century and is characterized by a strong vision of full inclusion for all members of society (Haualand & Holmström, 2018). Over the same period, Sweden also transformed from a monolingual and homogenous society to a much more multilingual and culturally diverse society; of Sweden’s approximately 10.5 million inhabitants, about 20% were born abroad (SCB, 2022). The deaf population consists of about 10,000 individuals (SDR, 2021), and in addition there are approximately 2,000 people who are deafblind (NKCDB, 2021). The Language Act states that the public society shall protect and promote Swedish Sign Language (svenskt teckenspråk [STS]) and that deaf people have the right to learn, develop, and use STS. This means that deaf people have the same legislated rights as other minority groups (Sami, Meänkieli, Finnish, Romani Chib, and Yiddish) or allophone language speakers to have access to interpreting services. Deaf people’s right to interpreters was affirmed in the 1994 Disability Reform (Swedish Government Official Reports [Statens offentliga utredningar;SOU], 2004:64). The right to use interpreters is also codified in several other Swedish laws. For example, the following laws all directly refer to interpreters: the Language Act (Swedish Code of Statutes [Svensk författningssamling; SFS],2009:600); the Health and Medical Services Act (SFS, 2017:30); the Support and Service for Persons With Certain Functional Impairments Act (SFS, 1993:387); the Administrative Procedure Act (SFS, 2017:900); the Code of Juridical Procedure (SFS, 1942:740); and the Administrative Court Procedure Act (SFS, 1971:201). Interpreting services are also codified, though more indirectly, in the Patient Act (SFS, 2014:821). We begin by contextualizing translation and interpreting of both signed and spoken languages in Sweden. We then talk about the status of STS and the professionalization of sign language interpreting (SLI) and discuss the development of interpreter education, with reference to deaf and hearing interpreters alike. We discuss the issue of readiness to work after formal education, as well as continuing professional development. Finally, we reflect on the status of STS/Swedish interpreting in the present day and on the current state of research into SLI and interpreter education.

The theory of language structure informs us about what we should expect when we want to investigate a certain construction. However, reality is often richer than what theories predict. In this study, we start from a theoretically informed set of hypotheses about the structure of wh-questions in sign language, we test them using a sign language corpus, a designed production experiment, and structured fieldwork in three sign languages, Swedish, Greek and French Sign Languages. The results will inform us on what type of contribution each research method can provide to reach accurate language descriptions.

The theory of language structure informs us about what we should expect when we want to investigate a certain construction. However, reality is often richer than what theories predict. In this study, we start from a theoretically informed set of hypotheses about the structure of wh-questions in sign language, we test them using a sign language corpus, a designed production experiment, and structured fieldwork in three sign languages, Swedish, Greek and French Sign Languages. The results will inform us on what type of contribution each research method can provide to reach accurate language descriptions.

This semi-systematic literature review examines signed language corpus research in the Nordic countries, with a quantitative and qualitative assessment of corpus research. The article first describes some critical components and functionalities of a signed language corpora. It then outlines the evolution of Nordic corpus research, highlighting Sweden’s pioneering role and subsequent developments in Finland and Norway. The findings suggest a progression from method-focused publications to those exploring linguistic phenomena within and across (signed) languages. Although the number of research publications is modest, there is a discernible shift towards comparative studies and applications in signed language teaching and learning.

Unsupervised representation learning offers a promising way of utilising large unannotated sign language resources found on the Internet. In this paper, a VQ-VAE model is trained to learn a codebook of motion primitives from sign language data. For training, we use isolated signs and sentences from a sign language dictionary. Three models are trained: one on isolated signs, one on sentences, and one mixed model. We test these models by comparing how well they are able to reconstruct held-out data from the dictionary, as well as an in-the-wild dataset based on sign language videos from YouTube. These data are characterized by less formal and more expressive signing than the dictionary items. Results show that the isolated sign model yields considerably higher reconstruction loss for the YouTube dataset, while the sentence model performs the best on this data. Further, an analysis of codebook usage reveals that the set of codes used by isolated signs and sentences differ significantly. In order to further understand the different character of the datasets, we carry out an analysis of the velocity profiles, which reveals that signing data in-the-wild has much higher average velocity than dictionary signs and phrases. We believe these differences also explain the large differences in reconstruction loss observed

The purpose of the interlingual index is to link the lexical resources of the sign languages of the project. This is the final project release of the index, which covers both the core sign languages of the project as well as additional sign languages.

Jordan Fenlon and Julie A. Hochgesang have done fantastic work in their collaboration on this long-awaited book on signed language corpora, which has been much anticipated by sign language researchers. This endeavor aims to propel conversations regarding the role of corpus linguistics as a methodology and its significance in sign language linguistics research. The book offers a practical and all-encompassing initiation into the application of corpus linguistics and corpus-based methodologies within the realm of sign language linguistics and applied linguistics, including sign language acquisition and teaching. The book unfolds across seven chapters authored by twelve well-known researchers and experts in their respective subfields. This collective expertise ensures a thorough exploration of the subject matter, making the book a valuable resource for those delving into the intricacies of signed language research and its practical applications in linguistic studies and language education.

The Multilingual Sign Language Wordnet is the first publicly available wordnet resource for sign languages. It is a growing multilingual resource providing data for eight sign languages to date. During the initial phase of its creation, the focus lay on producing the infrastructure to support various languages and to produce initial sets of content for them. This article represents the start of the second phase, in which the focus is moved to establishing overlapping coverage across the different sign languages. Building on the data produced so far, a new feature to assist annotation is introduced which leverages established partial synonymy between signs (inter- and cross-lingually) to discover likely additional synonymies. Other improvements to the annotation interface and workflow build directly on the experiences from the first phase. Working with the updated annotation interface, new data is produced for Polish Sign Language, Greek Sign Language and Swedish Sign Language.