Skip to content

i18n Human Reviewer (IHR) — Evolution Guide

Introduced in v1.7.0. i18n Human Reviewer is a docs-as-code persona with FCC phase Critique, Zachman cell USER/WHAT, and archetype The Interpreter. This guide distils the persona YAML into an actionable evolution path: where the role sits in its career track, the technical and behavioural milestones to hit in the first 180 days, a concrete 90-day onboarding playbook, the pitfalls most commonly encountered by IHR practitioners, and the escalation routes when the role reaches the edge of its charter.

Consult the complementary references:

Role snapshot

Field Value
Full title i18n Human Reviewer
Category docs_as_code
FCC phase Critique
Zachman cell USER/WHAT
Archetype The Interpreter
Introduced v1.7.0
Collaborators 5 (upstream: 2, peer: 1, downstream: 2)

Provides native-speaker review for LLM-generated translations produced by the v1.7.0 i18n pipeline. Owns the quality gate between Pillar-C LLM translation output (FR/ES/DE) and published localized docs, catches cultural-register drift, domain-vocabulary mistranslations, and style-guide violations that automated checks cannot detect.

Career Path

Where this persona sits. Technical translator → Domain i18n specialist → Human-in-the-loop translation reviewer. In FCC, IHR carries the Zachman cell USER/WHAT, which places the role in the user perspective; the column (what) indicates its primary column affinity with data, terms, and concepts. The role sits in the docs_as_code category and operates predominantly in the Critique phase of the FCC workflow.

Specialties that feed in. Native-level fluency in at least one target language (FR, ES, DE), domain literacy in AI/ML/governance vocabulary, technical translation style guides, CAT tools (memoQ, Trados, OmegaT), and LLM translation quality assessment. Practitioners typically arrive with 5+ years of translation experience before assuming the IHR role; the archetype The Interpreter recurs in several adjacent FCC personas (see ../archetype-atlas.md for the full archetype-to-persona map).

Advancement. Promotion to i18n Quality Lead (orchestrates IHR + pillar-c-translator) or cross-ecosystem localization steward. The conventional promotion signals for IHR are: (1) cited in three or more downstream role_collaborators lists; (2) repeated citation at the head of cross-reference traversal chains; (3) outputs consumed by at least two FCC phases. See ../evolution-pathways.md#champion-promotion-patterns for the full Champion promotion ladder.

Lateral moves. IHR shares archetype The Interpreter with other translation-oriented personas; a lateral move preserves the archetype identity while shifting category. Conversely, a move to a Champion role is one-way — once promoted the role acquires orchestration responsibilities on top of the archetype's core behaviour.

Skill Milestones

Milestones are stated as observable markers at four checkpoints. The 30/60/90 cadence maps to the standard FCC onboarding rhythm (foundational → structured → semantic); the 180-day checkpoint corresponds to the federated stage for this role.

30-day markers

  • Complete orientation to the i18n Human Reviewer role: read the persona YAML at src/fcc/data/personas/i18n_human_reviewer.yaml and the adoption checklist.
  • Review a sample 1,000-word LLM translation batch and produce a first quality-grading rubric.
  • Shadow at least one pillar-c-translator prompt cycle and one EDM ecosystem-docs cycle.
  • Identify the one discernment trait where your starting score is lowest and document a 60-day plan to improve it.

60-day markers

  • Deliver your first independent artefact: a target-language style guide with 50+ domain-term glossary entries.
  • Extend the existing responsibilities (5 total) with at least one new metric or heuristic you contribute back.
  • Demonstrate facility with CAT tools, markdown-aware diff review, and the Pillar-C LLM translation workflow; contribute a tool-chain improvement.
  • Attend one JV governance sync and one ecosystem integration review to understand cross-project touchpoints.

90-day markers

  • Own the docs/i18n/{fr,es,de}/ quality gate end-to-end (review queue + approval flow + published artefacts).
  • Publish a retrospective documenting what you learned about the five constraints in the persona spec.
  • Mentor the next adopter of the IHR role — pair-review their first artefact.
  • Refine your discernment-matrix self-ratings and request peer validation.

180-day markers

  • Carry full accountability for the IHR responsibilities list across at least one full release cycle.
  • Propose at least one refinement to i18n_human_reviewer.yaml based on observed gaps.
  • Represent the role in cross-ecosystem syncs and become the recognised escalation contact.
  • Score "scored" on all six discernment traits with differentiated rationale.

Core technical skills (from the persona YAML):

  • Native-level target-language fluency (FR / ES / DE)
  • Domain-vocabulary glossary curation (AI, ML, governance)
  • CAT tool fluency (memoQ, Trados, OmegaT) + markdown-aware diff
  • LLM-translation quality grading (accuracy, fluency, register)
  • Style-guide authoring + enforcement

First 90 Days

A concrete onboarding playbook for the first three months in the role.

Week 1 — Orient

  • Read the canonical YAML at src/fcc/data/personas/i18n_human_reviewer.yaml end-to-end. Pay particular attention to the R.I.S.C.E.A.R. block, the discernment matrix, and the constitution section inside doc_context.
  • Read ../evolution-pathways.md to understand where you sit in the four-stage maturity model.
  • Read ../archetype-atlas.md#interpreter for the archetype pattern.
  • Read the model card at docs/model-cards/ihr.md for risk classification and compliance context.

Weeks 2–4 — Produce the first deliverable

Your first deliverable should be aligned with the persona's primary expected output: a translation review bundle covering one target-language batch, with per-segment grades (accuracy, fluency, register), domain-glossary corrections, and a reviewer's summary.

Weeks 5–8 — Learn the toolchain

Develop working fluency with the following tools and techniques drawn from the persona's role_skills list:

  • Native-level target-language fluency (FR / ES / DE)
  • Domain-vocabulary glossary curation (AI, ML, governance)
  • CAT tool fluency (memoQ, Trados, OmegaT) + markdown-aware diff
  • LLM-translation quality grading (accuracy, fluency, register)
  • Style-guide authoring + enforcement

Pair with a seasoned Interpreter practitioner or external certified translator for at least two sessions to absorb idiom and discipline.

Weeks 9–12 — First review

By day 90 you should have:

  • Delivered at least two artefacts that pass the persona's adoption checklist.
  • Completed a peer review cycle with both upstream and downstream collaborators.
  • Documented a short retrospective on the five constraints in the persona YAML and how you honoured them.
  • Earned scored entries on at least three of the six discernment traits.

Adoption-checklist targets to hit by day 90:

  • 100% of merged translations have a human-reviewer signoff recorded in the commit trailer.
  • Domain glossary covers the top 200 terms across the translation surface.
  • Style guide published for each target language with example before/after pairs.
  • Zero unreviewed translations shipped to published docs.

Common Pitfalls

Failure modes specific to the IHR role, drawn from the persona's constraints, the Discernment Matrix, and the cross-reference patterns observed across v1.5.0–v1.6.2.

Role-specific pitfalls

  • Approving a translation that is grammatically perfect but culturally tone-deaf (register drift).
  • Missing inconsistent terminology where the same English term is translated differently across files.
  • Signing off on a translation that breaks mermaid diagram labels or front-matter YAML.
  • Accepting LLM output verbatim without probing the domain-specific edge cases (acronyms, license wording, governance terms).

Constraint-driven pitfalls

The following pitfalls are the negation of the constraints listed in the persona YAML. Violating any of them is a governance signal:

  • Constraint: No translation MAY ship without a human-reviewer signoff in the commit trailer.
  • Constraint: Every domain term MUST resolve to the canonical glossary entry.
  • Constraint: Reviewers MUST preserve markdown structure, front matter, and code-fence content unchanged.
  • Constraint: Register drift findings MUST be recorded in the style-guide revision log.
  • Constraint: LLM-translation grade MUST be logged with the batch ID for future-quality tracking.

Discernment-trait pitfalls

Pitfalls anchored in the Discernment Matrix — each one corresponds to a trait where the role must hold a high bar:

  • Humility drift: Defers ambiguous terminology to the domain subject-matter expert.
  • Professional Background drift: Deep native-language + domain fluency; fluent in CAT tooling.
  • Curiosity drift: Probes every new batch for novel terminology gaps.
  • Taste drift: Prefers clean, idiomatic target-language prose over literal translations.
  • Inclusivity drift: Welcomes regional dialect discussions; maintains neutral register where applicable.
  • Responsibility drift: Owns translation-quality integrity; audit-ready review log.

Escalation Points

When the IHR role reaches the edge of its charter, handoffs and escalations flow along the collaboration edges recorded in the persona YAML. The table below is read as: in this situation, hand off to this target with this artefact.

Collaboration graph

Upstream (this persona receives from):

  • pillar-c-translator — Receives LLM translation batches for human review.
  • MDC — Receives flagged diagram-embedded strings needing translation.

Peers (bidirectional):

  • EDM — Coordinates ecosystem-docs localization scope.

Downstream (this persona hands off to):

  • TLDR — Hands reconciled translations for three-layer integration.
  • RCRS — Flags corpus-refresh entries needing translation updates.

Escalation matrix

Situation Escalate to Artefact expected
Constraint violation observed in own output Forensic Auditor (FA) evidence packet with cited constraint + remediation plan
Downstream consumer reports drift from IHR artefacts TLDR reconciliation delta + remediation plan
Cross-project handoff required RHL (research) or IHL (innovation) handoff bundle with provenance metadata
Systematic LLM-translation defect across a batch pillar-c-translator prompt-tuning request + failing segment samples
Target language requires new native reviewer (beyond FR/ES/DE) Release Manager (RM) i18n expansion proposal + sourcing plan

Champion promotion. If you find yourself cited by three or more downstream personas' role_collaborators lists across consecutive releases, you may be an i18n Quality Lead-promotion candidate. See ../evolution-pathways.md#when-to-promote for the formal criteria.

See also