AI for Teaching in Sundanese

Say you teach a compulsory Basa Sunda class in a West Java primary school and want a short reading passage for next week. AI can draft one fast, but it's working with far less Sundanese-language training data than it has for Indonesian or English, so the draft needs a fluent reviewer before it reaches students.

Sundanese, spoken by tens of millions of people, is nonetheless what researchers call a low-resource language for AI, and that gap shapes every part of how these tools should be used in a Sundanese classroom.

Quick Answer: AI can help generate Sundanese vocabulary lists, reading passages, and worksheet structures, but its fluency in Sundanese specifically lags well behind Indonesian and English, since large language models train on far less Sundanese text. The most reliable workflow is generating structure and scaffolding, then having a fluent Sundanese speaker verify or supply the actual language content.

Sundanese (Basa Sunda) is spoken by an estimated 30 million or more people across generations, concentrated in West Java and Banten provinces — one of Indonesia's largest regional languages by speaker count, second only to Javanese. Indonesia is home to several hundred living languages, a total exceeded by only one other country worldwide, which makes "teaching in a regional Indonesian language" a genuinely different challenge from teaching in Indonesian (Bahasa Indonesia) itself.

For the wider regional picture, see AI in Education Around the World: A 2026 Regional Guide, and for AI-assisted entry-test prep in a different country's exam system, see AI for ECAT and Engineering Entry Tests.

What Makes Sundanese a Distinct Teaching Context

Sundanese instruction combines a genuinely large speaker base with a set of features that make generic AI tools less reliable than they are for English or Indonesian content.

Tens of Millions of Speakers, Still Low-Resource for AI

Low-resource, in AI research, doesn't mean few speakers — it means comparatively little digitized text available to train a model. Sundanese has tens of millions of speakers but a fraction of the written online content that Indonesian or English has, so a model's fluency, idiom, and vocabulary range in Sundanese specifically trail well behind its Indonesian-language output.

Latin Script Today, Aksara Sunda as Cultural Heritage

Modern Sundanese instruction is written almost entirely in Latin script, the same alphabet used for Indonesian and English. Aksara Sunda, a traditional script with roots in older Brahmic-derived writing systems of the archipelago, survives today mainly as cultural heritage content — street signage in parts of West Java, ceremonial contexts, and dedicated heritage units — rather than as the everyday classroom writing system.

Speech Levels: Undak Usuk Basa

Like Javanese, Sundanese traditionally uses a system of speech levels (undak usuk basa) that shifts vocabulary choice based on the relative age or social standing of speaker and listener — a more formal register for addressing an elder, a more casual one among peers. This is a genuinely difficult feature for AI to get right, since register choice depends on social context a text prompt alone often doesn't specify.

West Java's Muatan Lokal Mandate

Sundanese-language instruction in schools isn't optional in much of its home region — it's built into provincial policy.

What Muatan Lokal Actually Requires

Muatan lokal ("local content") is Indonesia's policy mechanism for province- or region-specific compulsory subjects alongside the national curriculum. West Java's provincial regulations designate Sundanese-language instruction as a compulsory muatan lokal subject in schools across the province, which is why Basa Sunda appears as its own timetabled subject rather than an elective.

How This Fits Within Kurikulum Merdeka Nationally

Indonesia's Kurikulum Merdeka, rolled out nationally from 2022 by the Ministry of Education, Culture, Research, and Technology (Kemendikbudristek), gives schools more flexibility in how they sequence content — muatan lokal subjects like Basa Sunda sit alongside this national framework rather than being replaced by it. For a fuller look at Kurikulum Merdeka planning itself, see Creating Kurikulum Merdeka Lesson Plans (Modul Ajar) With AI.

Why Intergenerational Transmission Is Worth Naming

Even with tens of millions of speakers and a compulsory-subject status, sociolinguistic observation in West Java has repeatedly noted a generational shift, particularly in urban centers like Bandung, toward Indonesian as a first home language among younger families. Badan Bahasa, Indonesia's national language body under Kemendikbudristek, runs regional-language revitalization programs partly in response to this pattern — worth knowing, since it shapes how much existing Sundanese fluency a teacher can assume in an urban classroom versus a rural one.

Where AI Realistically Helps — and Where It Needs a Human Check

Matching AI's actual strengths to a Sundanese classroom means being honest about where the low-resource gap shows up most.

Vocabulary and Reading Material Generation

AI is most dependable generating structure: a vocabulary list template, a reading-comprehension question set format, a worksheet layout — even when the actual Sundanese words and sentences inside that structure need a fluent teacher's input or review.

Where AI Output Needs a Fluent Reviewer

Register (undak usuk basa), idiom, and natural word order are the highest-risk areas for AI-generated Sundanese content, precisely because these are the parts most dependent on cultural and social context a model can't infer from a short prompt alone.

Why Generic AI Chatbots Struggle Specifically With Sundanese

Understanding why the low-resource gap shows up helps explain which prompting strategies actually close it and which don't.

Training Data Volume Drives Fluency, Not Speaker Count

A language model's fluency in any given language tracks roughly with how much digitized text in that language it was trained on — news articles, books, forums, government documents. Indonesian has a large, growing digital footprint; Sundanese, despite its tens of millions of speakers, has considerably less, since much everyday Sundanese communication historically happened orally or in contexts that never got digitized.

Dialectal Variation Adds a Further Layer

Sundanese itself varies across sub-regions — the variety spoken in Bandung differs somewhat from that in Priangan's more rural areas, or from the Banten periphery. A model trained predominantly on whatever Sundanese text exists online may lean toward one variety without a teacher realizing it, which matters if local classroom usage differs from what gets generated.

What Actually Helps: Explicit Constraints

Naming the exact vocabulary level, topic, and — where relevant — regional variety in a prompt narrows the model's guesswork considerably. "Simple Sundanese, common Bandung-area vocabulary, Grade 2 level" produces a more usable draft than "write this in Sundanese" alone, even though both still need a fluent check.

Step-by-Step: Building a Sundanese Lesson With AI Support

1. Decide what needs to be generated in Sundanese versus what can stay in Indonesian or English. Instructions, question formats, and rubrics can often stay in a higher-resource language even when the reading passage itself is Sundanese.
2. Ask for a worksheet or passage structure first, specifying grade level, topic, and length — this scaffolding step is where AI is most reliable.
3. Generate a draft vocabulary list or simple-sentence passage in Sundanese, using common, everyday words rather than requesting idiomatic or highly formal language.
4. Specify the intended speech level explicitly if generating dialogue — casual peer-to-peer, or a more formal register — even though the output still needs a check.
5. Have a fluent Sundanese speaker — a colleague, an aide, or your own judgment if you're a fluent speaker — review every generated sentence before it reaches students.
6. Reserve Aksara Sunda script content for dedicated heritage units, sourced from verified cultural or educational references rather than AI generation.
7. Save verified passages and vocabulary sets into a running bank, since rebuilding verified content from scratch every lesson is the slowest part of this workflow.

A Worked Example

Say you're building a Grade 3 Basa Sunda unit on family vocabulary. A teacher could ask AI for an English-language worksheet structure — a labeling diagram, a matching exercise, a short-answer section — then separately generate a simple candidate vocabulary list (bapa, indung, lanceuk) in Sundanese for a fluent reviewer to confirm before it's dropped into the worksheet.

Pro tip: Keep a running, reviewer-verified glossary organized by topic (family, numbers, classroom objects, colors). Once a term is checked once, it doesn't need re-verifying every time it resurfaces in a new worksheet.

Combining Sundanese Content With Broader Literacy Skills

A Sundanese reading passage still teaches the same underlying comprehension skills — finding the main idea, making inferences, building vocabulary in context — that any reading instruction targets, regardless of language. Once a passage is verified, the same tiered-question approach used in any other language classroom (literal, inferential, evaluative questions) applies directly, which means the comprehension-question generation step doesn't carry the same low-resource risk the passage's own wording does.

Handling Mixed-Proficiency Classrooms

A Sundanese classroom in an urban center like Bandung increasingly includes learners with uneven home exposure to the language, alongside others who speak it fluently at home. Generating two versions of the same worksheet — one with more vocabulary scaffolding, one without — is a low-effort way to differentiate without redesigning the whole lesson, once a base version has been verified.

Rural West Java classrooms tend to see the opposite pattern — stronger home-language fluency but less exposure to formal, written Sundanese register, since everyday use skews conversational. Differentiation here often runs the other direction: building comfort with formal written structure rather than basic vocabulary recognition.

Tools & Technology Comparison

EduGenius's flashcard and worksheet formats work well as a structural scaffold for Sundanese vocabulary practice — a teacher can set up the format and question types quickly, then populate the Sundanese-language content directly or drop in verified terms from a reviewed glossary.

What This Looks Like Across Grade Levels

Across every level, the pattern holds: AI's contribution shrinks as the content gets more linguistically specific to Sundanese, and grows again on the structural, format, and organizational side of lesson-building. Planning around that pattern deliberately — rather than expecting uniform AI reliability across every part of a lesson — is what makes the workflow actually save time instead of creating a new verification burden.

Building Cultural Context Alongside Language

Sundanese instruction rarely stops at vocabulary and grammar — much of what makes it meaningful to learners is the cultural context wrapped around the language itself.

Folklore and Oral Tradition

West Java's oral storytelling tradition — including well-known folk narratives passed down regionally — offers rich, culturally grounded reading material. AI can help draft a simplified, grade-appropriate retelling structure, but the actual story content and cultural details are best sourced from or verified against community knowledge and published regional folklore collections, not generated freely.

Traditional Arts and Vocabulary Connections

Sundanese vocabulary connects naturally to regional arts — angklung (a traditional bamboo musical instrument), wayang golek (wooden rod-puppet theater), and traditional dance forms — giving a vocabulary unit a cultural anchor beyond a flat word list. AI can generate a structured vocabulary-and-context worksheet template around one of these topics quickly, with the specific cultural facts checked against a reliable source.

Why This Matters for Engagement, Not Just Accuracy

A vocabulary list divorced from any cultural context tends to feel like an arbitrary memorization task to students, while the same words anchored to a familiar cultural reference — a song, an instrument, a local tradition — give learners a reason to care about the term beyond the test. This is a genuinely underused AI application: generating the worksheet structure around a cultural theme a teacher specifies, freeing time to focus on sourcing the content accurately rather than building the format from scratch.