Course Dependency and Terminology Inventory

Translate the learner action, not just the lesson text

Before translating an overseas course for China, prepare a proposed shared record of the terms that cannot drift, the claims that cannot be strengthened, and the dependencies behind each learner action. The inventory becomes a controlled source record only after the creator or designated reviewer approves it; a dependency audit is the risk analysis it enables.

Most courses reveal their true complexity in the exercise, not the video. A lesson may be ten minutes long but require an account verification email, a cloud template, a model subscription, a shared database, an API key and a live support handoff. Those are not production details. They are part of the learning product. But before anyone diagnoses or replaces them, the team needs to know what the creator actually called the method, what outcome was promised, and which example is non-negotiable.

The point of an audit is not to prove that every foreign tool fails in China. That would be untrue and unhelpful. The point is to replace a vague question—“Can we localize this?”—with a list of testable dependencies.

This is not the published dependency ledger

OriBridge’s published Localizing an Expert Product for China is a reader-facing dependency ledger: it maps learner actions to access, account, asset and support risks. This article is different. It is the pre-edit inventory that preserves terms and source decisions so a translator, a subject reviewer and a delivery operator do not silently solve three different problems.

Do not publish both pages unless the reader can see that difference. The ledger asks, what may break? The inventory asks, what exactly are we allowed to change, translate, replace or leave unresolved?

The five-tab worksheet

Use this as a proposed internal format, not as evidence that a creator has approved a translation. Each row needs an owner and a version, so a later editor can see whether a Chinese term, example or tool replacement is approved, pending or prohibited.

E0 means captured from the source course; E1 means written creator or designated-reviewer approval; E2 means a re-openable China-side official rule; E3 means a controlled, dated route test. A field without its required evidence tier is pending, not approved.

1. Tool: what does the learner have to operate?

Write the named service, exact action, account type, device and the learning outcome it serves. Do not write merely “AI tool.”

If a course is intended for a WeChat Store education route, the current category rules flag content involving foreign account services, VPN/circumvention and interface-call topics among relevant restrictions. The source is WeChat Store’s “微信小店教育培训类目管理规则”, revised 2026-04-07 and checked 2026-08-12. That rule does not answer general access to the tool. It tells the editor to treat the tool instruction and the commercial route as separate risks.

2. Asset: what must the learner receive or reuse?

List every workbook, template, screenshot, recording, third-party image, sample dataset and linked repository. For each, record the rights holder, allowed use, version and fallback.

An older marketplace-import note was not retained here as a re-openable public URL, so it is not used as a factual basis. The only safe general conclusion is procedural: before an asset crosses any distribution route, record the asset, proposed route, owner, version and unresolved test. Do not infer a workaround, eligibility or a delivery result from an old help-centre note.

3. Identity: what must the learner sign up for or disclose?

Record each login, verification message, group invitation, consent step and access role. A learner who cannot receive a confirmation or enter the right workspace has not started the lesson, no matter how good the translated script is.

This is also where teams must distinguish product records from informal messages. Do not solve a broken account journey by asking learners to send credentials or personal data through ungoverned channels.

4. Promise: what does the lesson make the learner believe they can do?

Compare the title, cover, learning outcome, examples, CTA and promoter script. If the source course promises client acquisition, a route may impose restrictions that its original market did not. If it promises a result that depends on a blocked or unapproved tool, the promise must be narrowed, redesigned with creator approval, or withdrawn.

5. Recovery: what happens when one step breaks?

Write an owner, a response window, a fallback and a stop condition. “Contact support” is not a recovery plan unless someone knows which support team owns the underlying failure.

The audit’s best outcome is sometimes a stop decision: do not translate the lesson until the essential dependency has an authorized, testable replacement.

The inventory fields that make a terminology record useful

For each selected lesson, add three fields a simple access audit misses:

• source term and approved Chinese rendering: retain the original term, the current Chinese term, owner and version; do not invent a new technical label merely because it is shorter;
• immutable teaching claim: write what must not become stronger or broader in translation—for example, “may help decide what to test” must not become “will produce results”;
• change authority: identify whether the creator, subject reviewer or operator may change an example, tool name, screenshot or learner instruction.

This is the boundary between localization work and unauthorized rewriting. The table is a recommended record structure; this article currently has no public creator-specific terminology or approval artefact, so it makes no claim that any particular glossary has been approved.

Hotspot-Demand Card: the stable problem behind “just translate it”

An illustrative completed row

This is an illustrative method, not an actual test log. Do not mark “works in China” until the defined path has been tested under recorded conditions.

Add a model/version row when the tool is part of the lesson

If a named AI model appears in a lesson, add a row for the model or version, the visible example, the learner action, the public-facing or internal service scope, the change authority, and the fallback. The Chinese generative-AI service rules are a reason to keep that boundary explicit; they are not a shortcut to a tool-availability or course-approval conclusion. Recheck the official Measures for the Administration of Generative Artificial Intelligence Services when the service, audience, or public-facing promise changes.

Mark the row as an internal control record, not as a legal checklist. A model name in a screenshot, a successful internal rehearsal, and a learner’s ability to complete the promised task are three different facts.

Next step

Read the localization-debt dependency ledger first if you need to identify what may break. Use the MCP lesson reproducibility record when the dependency is an AI workflow that must be repeatable rather than merely translated. Neither article gives permission to alter a creator asset.

Use the inventory before requesting validation. Start with one representative lesson and mark each row captured, pending approval, approved, needs route test or do not use. When the course owner has approved the proposed changes and the essential route is testable, request a scoped China validation review rather than commissioning a full translation blind.

China Evidence Card

Why this is hard to obtain from an English AI: the audit links Chinese route rules and course-asset handling to the exact learner action, rather than treating access as a generic country label.

OriBridge’s added judgement: classify dependencies by the learner action and promised outcome; do not choose a local substitute until you know what the original component was actually doing.

Public boundary: this is an operational template, not a completed evidence record, legal opinion or technical access advice. It does not certify permissions, availability, privacy, payment, compliance or learner results. The three platform/rule sources establish dependency questions only; they do not supply a creator-specific glossary or approval. Any team using this template must attach its own written creator/designated-reviewer approvals before treating a proposed term, claim or change as approved.

Related reading

• Localizing an Expert Product for China: What Must Change Beyond Translation
• End-to-End Access, Payment, and Delivery Test Script