AfricaSign -- A Crowd-sourcing Platform for the Documentation of STEM Vocabulary in African Sign Languages
Abdelhadi Soudi, Kristof Van Laerhoven, Elmostafa Bou-Souf · 2019 · Proceedings of the 21st International ACM SIGACCESS Conference on Computers and Accessibility (ASSETS) · doi:10.1145/3308561.3354592
Summary
This demonstration paper presents AfricaSign, a web-based crowdsourcing platform designed to enable African Deaf communities to document their endangered sign languages, with a focus on early-grade and STEM vocabulary. Most African sign languages are severely under-studied and endangered, yet there has been minimal technological support for their documentation. The platform offers three input modes for contributing signs: uploading an existing video, recording a sign using the device's camera, or describing the sign through a signing avatar interface where users specify manual markers (hand shape, location, movement, orientation — the four cheremes) and non-manual markers (facial expressions) through pictorial drop-down menus. The avatar-based input mode is novel in using a user-friendly chereme-based strategy to animate signs, building on established notation systems like HamNoSys and the SiGML markup language from the EU ViSiCAST and eSIGN projects. Documentation is organized by country and region to accommodate the significant regional variation that exists within sign languages. Users register with demographic information including country, region, and deafness affiliation (Deaf themselves, CODA — children of Deaf adults, etc.). The target vocabulary is based on an enhanced version of the MacArthur-Bates Communicative Development Inventory. The platform supports Arabic, English, German, and French interface languages.
Key findings
The platform was deployed in a two-phase approach. Phase I involved restricted access with partners from five African countries (Ethiopia, Ghana, Kenya, Ivory Coast, and Morocco), each with a country moderator coordinating with regional collaborators to identify contributors. This phase served as a testing protocol with quality assurance mechanisms. Phase II planned unrestricted access to expand the contributor network across the African continent. The research addresses both linguistic and technological questions: linguistically, it aims to uncover typological features of African sign languages, understand the nature and degree of regional variation, and explore whether documentation can help address negative attitudes toward Deafness in Africa. Technologically, it tackles how to design signing avatars suitable for a heterogeneous set of languages and cultures, how to evaluate avatar quality in the absence of standardized protocols, and how to effectively use crowdsourcing for sign language data collection in African contexts.
Relevance
This work addresses a critical intersection of language preservation, Deaf education, and digital accessibility in the developing world. STEM vocabulary documentation is particularly important because the lack of standardized STEM signs in many African sign languages creates a direct barrier to education — Deaf students cannot access mathematics, science, and technology concepts when there are no signs for the relevant terminology. By enabling communities to document their own signs, AfricaSign takes a bottom-up approach that respects linguistic diversity rather than imposing signs from dominant sign languages like ASL or BSL. The crowdsourcing model is well-suited to this challenge because sign languages are community-owned resources, and the Deaf community members themselves are the primary authorities on their language. For accessibility practitioners, the platform demonstrates how technology can serve both documentation (preserving endangered languages) and education (creating shared STEM vocabulary resources) simultaneously. The avatar-based input mode is particularly innovative for contexts where video recording may be impractical due to bandwidth limitations or privacy concerns common in African internet environments. The regional variation accommodation is a thoughtful design decision that reflects the linguistic reality that sign languages vary significantly even within a single country.
Tags: sign language · Africa · endangered languages · crowdsourcing · signing avatar · STEM accessibility · Deaf education · language documentation · digital inclusion · developing nations
Standards referenced: HamNoSys · SiGML