Implemented resources include the international portal, the Resource Creator and V-book builder, Video Classroom, and institutional administration.

Potential areas for continued improvement include multilingual resource libraries, V-book creation and publishing, collaborative video classrooms, and institutional management tools. These improvements are possibilities for future development, not newly available features.

Includes searches by place, age group, gender and category; access to annotated videos, corpus examples and lexical cross-references.

Further development may improve corpus search, access to annotated signing videos and sociolinguistic metadata, linguistic filters, and links between corpus examples and SignBank lexical entries. Integration, publication and testing with Deaf communities remain subject to further work and validation.

The existing prototypes demonstrate aspects of corpus exploration and linguistic data access. Completing the integration with SignBank, preparing the resource for wider availability, and conducting further technical, linguistic and community-based evaluation are separate development steps.

Proposed changes include responsive design, more accessible navigation and cross-platform links without limiting specialised SignBank functionality.

A future integration may connect navigation between SignBank dictionary entries, corpus examples, linguistic research resources and educational tools within the VisuoLab ecosystem. The scope may include interface improvements, contextual links and coordinated access to related resources. Data exchange, shared user accounts, authentication and permissions would require separate technical assessment and agreement between the relevant partners before being included in the integration scope.

Accessibility assessment, implementation and verification: The initiative distinguishes three stages: (1) assessing accessibility through technical and user-centred evaluation to identify barriers; (2) implementing improvements based on the findings and agreed priorities; and (3) verifying the implemented changes through subsequent technical and user testing. The proposed scope focuses on the assessment stage. Implementation and verification are separate activities to be planned and agreed with the relevant partners.

Proposed assessment deliverables: The assessment is expected to produce (1) a documented scope identifying the pages, interfaces and user journeys examined; (2) a description of the evaluation methods, including automated checks, manual review and user-centred testing; (3) a record of identified accessibility barriers, with supporting evidence and their effects on user interaction; and (4) a prioritised set of recommendations to guide future improvements. The final scope, methods and deliverables will be defined with the participating partners before the assessment begins.

User participation and access conditions: The proposed assessment should consider participation by Deaf sign language users, people with different disabilities, educators, researchers and other relevant user groups. Evaluation activities should account for participants’ language preferences, communication needs, assistive technologies and access requirements. Appropriate arrangements for accessible instructions, participation and feedback should be planned with the partners and participants involved.

Accessibility standards and conformance: WCAG 2.2 AA and EN 301 549 are proposed references for the accessibility assessment. Their use as evaluation criteria does not establish that VisuoLab has achieved conformance or received certification. Any future conformance statement would require appropriate evaluation evidence, a defined scope and separate validation.

AI research directions: The proposed initiative brings together four distinct research directions: (1) sign recognition, investigating how signs and linguistic patterns might be identified in video; (2) sign-to-text, exploring assisted conversion of signed video into textual representations; (3) multimodal interaction, studying ways to connect signing videos, text and linguistic references in interactive systems; and (4) avatar-based sign language generation, investigating the possibilities and limitations of digitally generated signing. Supporting research activities include data and infrastructure, recognition, multimodal models, avatar research and international validation.

Research hypotheses and limitations: These research directions explore whether AI methods can support linguistic documentation, video analysis, educational access and interaction with sign language resources. Their feasibility and usefulness remain to be investigated through research and evaluation. Results may be affected by the availability and quality of suitable data, linguistic variation, context, uncertainty in model outputs and limitations of current technologies. Research prototypes should be evaluated for linguistic, cultural, technical and ethical adequacy, with human oversight and opportunities for correction. No particular level of accuracy, translation capability or production-ready service is assumed.

Data rights, licensing and consent: Any proposed use of SignBank lexical resources, corpus videos, annotations or other materials for AI training or evaluation should be preceded by a review of the applicable licenses, permissions, consent conditions and agreements with data contributors and relevant institutions. Access to a resource for linguistic documentation or research does not automatically authorize its use for training AI models. Decisions about dataset selection, permitted uses, access restrictions and any necessary additional permissions should be established with the relevant partners before such activities begin.

Linguistic and cultural evaluation: Research evaluation should consider linguistic variation, regional and community differences, signing context, non-manual features and the cultural appropriateness of AI-generated or AI-assisted outputs. Evaluation methods should involve relevant linguistic expertise and Deaf community participation. Research should also examine uncertainty, error patterns and the potential consequences of incorrect recognition, textual representations or generated signing, particularly in educational and research contexts. Findings, limitations and unresolved risks should be documented to inform decisions about further development and appropriate uses.

Human review, correction and appropriate use: AI-assisted outputs should be treated as research results or provisional suggestions requiring human interpretation and review, rather than authoritative linguistic representations or validated translations. Proposed workflows should allow qualified reviewers, including relevant linguistic specialists and Deaf sign language users, to examine results, identify errors, provide corrections and document unresolved uncertainties. The level of review and the suitability of any output should be assessed according to its intended context of use, particularly in linguistic research, educational materials and public-facing applications. Human review procedures and responsibilities would need to be defined and evaluated before wider deployment.

Technical development and integration planning: The initiative may explore AI-assisted workflows connected to the VisuoLab V-book editor, SignBank resources and relevant educational references. Potential components include specialised AI assistants, integrations with linguistic resources and tools for pedagogical planning, sign language scripts, visual composition and accessibility guidance. The architecture, number and responsibilities of agents, MCP servers, data flows and integration requirements remain matters for a separate technical proposal and evaluation with the relevant partners. Pilot activities, educator training and impact evaluation may be considered as part of an agreed development scope.

Curricular references and local adaptation: For proposed educational pilots in Brazil, the development of bilingual teaching materials may consider the Base Nacional Comum Curricular (BNCC), together with relevant local curricula, educational practices and the linguistic and cultural contexts of Deaf learners. The applicability of curricular references, pedagogical methods and language resources should be assessed with participating educators, Deaf communities and institutional partners. Implementations in other countries or educational contexts would require adaptation to their own curricular frameworks, languages and pedagogical priorities rather than assuming the BNCC as a universal reference.

AI-assisted drafting and human-led signing: The proposed V-book AI initiative may assist educators in preparing and revising bilingual teaching materials through pedagogical suggestions, glossary drafts, sign language script proposals, visual composition guidance and accessibility recommendations. These AI-generated contributions are intended as editable drafts and supporting resources, not finished or automatically validated educational content. The selection of linguistic forms, production of signed videos and validation of sign language content require appropriate human expertise, including Deaf sign language users and relevant educational and linguistic specialists. The initiative does not assume automatic generation or linguistic validation of signed videos.

Data flows, service providers and hosting: Before introducing new AI services or integrations, the proposed technical assessment should examine which information may be exchanged between VisuoLab, SignBank and any external systems; which providers and hosting arrangements would be involved; and how access, storage, retention and deletion of data would be managed. Security, privacy, permissions and applicable institutional requirements should be considered alongside technical feasibility.