Indigenous Language Certification Goes Local in August: A Digital Service Index for Registration, Testing, Same-Day Certification and AI Reminders across Taiwan’s 55 Indigenous Townships

Council of Indigenous Peoples, Taiwan announced on August 19 that Indigenous Language Proficiency Certification will introduce local testing beginning in August 2026. The model includes both school-based and township-based testing. Thirty school-based sessions are planned, while six township-based sessions are scheduled from August 29 through November in Hualien, Pingtung, Lanyu, Kinmen, Taitung and Penghu for beginner and intermediate candidates. The reform also introduces a same-day sequence of testing, scoring and certificate issuance, allowing candidates who complete a morning session to receive a certificate as early as that afternoon for use in further education, employment and qualification procedures. The Council also states that testing guides will be distributed to relevant local agencies and the offices of all 55 Indigenous townships, towns, cities and districts. See Council of Indigenous Peoples, Taiwan | Local Indigenous Language Certification Testing.

The practical lesson for the 55 Indigenous townships is that a public service is moving from “people must travel to find the system” toward “the system moves closer to people.” Township offices, schools, language teachers and community organizations can use this change to consolidate information now scattered across news releases, testing guides, examination websites, foundation websites and official messaging channels into one trusted entry point: which test applies, which level it serves, when registration opens, where the test is held, where the authoritative source is, when the certificate becomes available and whom to contact when information is unclear. Once this basic data governance is in place, RAG, chatbots and reminders can reduce search friction instead of creating another page that staff must separately maintain.

Layer 1: Turn certification information into service cards so residents can reach the correct entry point in a few steps

A common service-card schema could include `service type`, `level`, `applicable area`, `test date`, `registration period`, `test location`, `pre-registration requirement`, `certificate delivery`, `official source`, `last review date` and `service contact`. When a session has not yet published its detailed schedule, the card should explicitly say “awaiting official announcement” rather than allowing AIGC to infer a date, quota or venue. The Council already points users to its own website, the Indigenous Languages Research and Development Foundation, the local-testing website, an official LINE account and customer-service numbers. A township index should help residents reach those authoritative channels more quickly rather than replace them. See the 2026 Local Indigenous Language Certification Testing site and the Indigenous Languages Research and Development Foundation.

A homepage or chatbot could first ask three simple questions: “Where are you located?”, “Are you looking for beginner or intermediate level?”, and “Do you want registration, testing or certificate information?” The system then answers only from official records that are still valid. For elders and people who do not prefer conversational interfaces, the same service should keep large-text cards, one-tap calling and downloadable guides. Schools and tribal colleges can additionally use date reminders for classes or groups. The aim is a multi-entry public service, not a requirement that every resident learn to use AI before receiving information.

Layer 2: AI can assist retrieval and reminders, while official status is controlled by structured data

The new same-day testing, scoring and certification process can be represented as a clear service-state sequence: `not yet open → registration open → registration closed → awaiting test → tested → result/certificate available`. Before generating an answer, a RAG service can filter by location, level, date and active status. If a record is expired or two official sources do not agree, the system can display both source dates and route the item for human review. This helps prevent a chatbot from using last year’s testing guide to answer this year’s candidate.

The Indigenous Languages Research and Development Foundation already maintains an Indigenous-language AI results website presenting tools and outcomes related to speech recognition, speech synthesis and other language technologies, together with channels for feedback. These tools can support learning, practice or assisted transcription, but their outputs should not be presented as official certification scores. Formal certification remains governed by the testing system and results announced by the responsible authorities. See the Foundation’s Indigenous Language AI results site. If a township adds pronunciation practice, language prompts or a learning portfolio, it should keep that “learning assistance” module visibly separate from the “official registration and certification” module.

Layer 3: Connect one testing notice to a year-round Indigenous-language learning service chain

A local-testing notice loses much of its long-term value if it appears only as a poster before the test. A more useful design connects the certification service card to language courses, teachers and language promoters, learning materials, speaking practice, examination rules and explanations of certificate uses. When someone asks, “I want to take intermediate Amis,” the system can show the currently valid official testing information and approved learning resources or local courses. It should not independently determine eligibility or promise that a certificate automatically satisfies every education or employment rule; those decisions should link back to the relevant authority.

Beginning in 2027, the Council plans to hold the nationwide Indigenous-language certification examination once each December, with the highest level held every two years, while gradually making local testing a regular service. That means the local digital service must support changing rules. Each record should carry its year, version and last-review date rather than permanently embedding the 2026 schedule in chatbot knowledge. A simple safeguard is to create a new version when each year’s guide is released, move the old version to historical reference status and allow the public-facing RAG service to retrieve only records active for the current year by default. See the Council of Indigenous Peoples, Taiwan announcement.

Layer 4: When language data enters AI, provenance, permissions and intended use should travel with it

Language-learning and certification services will increasingly generate text, recordings, teaching materials, example sentences and user feedback. These records should not automatically become freely reusable model-training data simply because they may help AI. UNESCO’s August 2026 event, “Building Data Commons for Indigenous Languages and Cultures,” explicitly asks how Indigenous Peoples can retain control over data about their languages and cultures, while discussing Indigenous data sovereignty, the CARE Principles and social licenses. See UNESCO | Building Data Commons for Indigenous Languages and Cultures.

For the 55 Indigenous townships, permissions for `public viewing`, `teaching use`, `RAG retrieval`, `AIGC rewriting` and `model training` can be treated as separate fields. Material visible on a public website does not automatically authorize every AI use; recordings contributed by a teacher or community member should retain provenance, permitted uses, reuse limits and withdrawal procedures. The Global Indigenous Data Alliance | CARE Principles emphasize Collective Benefit, Authority to Control, Responsibility and Ethics, while Local Contexts provides TK Labels and related mechanisms for expressing cultural authority, provenance and conditions of use. These principles can become executable metadata in a RAG system rather than remaining only as an ethics statement on a website.

Layer 5: Begin with a 90-day MVP that the 55 Indigenous townships can share and adapt

During the first 30 days, focus only on trusted data: map the Council, the Foundation, the local-testing website and relevant local notices into a common schema, with a source and reviewer for every record. During the next 30 days, build a mobile-first service index and test whether location, level, dates, phone numbers, registration links and “awaiting announcement” states are clear. During the final 30 days, add chatbot/RAG retrieval and reminders, then test them with real questions such as “When is the intermediate test in Taitung?”, “Where do I register for the Lanyu session?”, “If I miss this session, is there another one?”, and “Can I receive the certificate today?” Whenever the authority has not yet published the information, the system should say that it remains pending and provide the official source instead of guessing.

Useful performance measures include how many steps residents need to reach the correct entry point, whether expired records automatically leave the active index, how quickly reported errors are corrected, how much language-accessible and easy-to-read information is available, whether paper and digital channels remain consistent, and whether a chatbot routes uncertain questions to an official service contact. These indicators reveal more about public-service quality than page views alone and allow the 55 Indigenous townships to share a common data structure while retaining local courses, language needs and community-specific services.

For Yuan Media AI’s 55-township announcement index, Chatbot MVP and Two-Eyed Seeing Lab, this certification reform offers a practical model: first turn official public services into traceable, updateable and expirable data; then use AI to help residents find them. When Indigenous-language data is later used by AI, govern community authority and purpose at the same time. This can help residents access testing while turning a one-time announcement into a maintainable Indigenous-language digital public service.

Sources

• Council of Indigenous Peoples, Taiwan | Local Indigenous Language Certification Testing
• 2026 Local Indigenous Language Certification Testing Website
• Indigenous Languages Research and Development Foundation
• Indigenous Language AI Results
• UNESCO | Building Data Commons for Indigenous Languages and Cultures
• Global Indigenous Data Alliance | CARE Principles
• Local Contexts | Grounding Indigenous Rights