Let Books Localization Case Study

Wiki English

Let Books is a living case study in building a multilingual knowledge platform and product vision before a full backend application exists.


Why it is useful

It shows how a project can:

  • define locale scope early
  • keep accessibility tied to localization
  • use English-first authoring without collapsing into English-only publishing
  • document governance before scale arrives

Slovenian AI-review example

The repository includes a concrete Slovenian review example where the AI draft preserved broad meaning but still required native-speaker correction for modality, fluency, and policy-register wording.

The detailed record now lives in the Native-Speaker Review Findings section of ../localization-at-scale-program.md and in ../style-guide/localization/ai-translation-review-records.md.

Key lessons:

  • grammar errors can survive even when the sentence seems understandable
  • modality is especially vulnerable to literal translation
  • policy and product-spec language often needs domain-specific register rather than dictionary-level equivalence
  • automated checks rarely catch subtle fluency and modality issues on their own

Mixed-language publishing example

Source article: docs/blog/sl/the-cost-of-english-only-software.md

This article family provided a concrete example of a localized title and partial localized shell coexisting with English reader-facing publication surfaces.

The detailed program-level interpretation now lives in the Native-Speaker Review Findings section of ../localization-at-scale-program.md and in ../localization-audit-report.md.

Repository-specific lessons:

  • coverage is not the same as completeness
  • source metadata matters because summaries can leak into publication surfaces
  • taxonomy localization matters because English topic labels are visible reader-facing defects
  • generated HTML should be validated, not trusted blindly

Reusable benchmark examples

Structured review examples should also be stored as reusable benchmark examples so future LLM evaluation can measure:

  • grammar reliability
  • modality handling
  • terminology precision
  • policy-register accuracy
  • quality of reviewer rationale capture