Validation summary

Everything marked during the session. Export it and attach it to the requirement register. Feedback is saved in this browser only.

Step 24 of 24 · Presenter note
Everything marked during the session, ready to export into the requirement register.
0Confirmed
0Need a change
0Open questions
149Not reviewed, of 149

Workshop

All requirements and questions

ItemRequirement or questionStatusNote
Step 1 · End-to-end flow
H-01
Requirement
One platform with a common translation, terminology, quality, review and audit layer; format-specific connectors handle extraction and reconstruction. §7.1Not reviewed
H-02
Requirement
Phase 1 covers the Windchill flow end to end: request, pre-flight, GTL approval, DITA/XLIFF processing, TM, AI, QA, proofreading, Windchill return and SAP/PNC publication. §29Not reviewed
H-03
Requirement
Independent requests, Office/HTML/PDF formats, Articulate, SuccessFactors and FieldCheckEditor follow in Phase 2. §29Not reviewed
H-04
Requirement
Windchill topic reuse and TM segment reuse are measured separately. §7.3Not reviewed
HQ-01
Question
Is Phase 1 correctly limited to the Windchill flow, with SAP/PNC as the only other destination? §29Not reviewed
HQ-02
Question
Should Acolad be phased out language by language (per wave) rather than on one date? §32Not reviewed
Step 2 · Discovery cockpit
P0-01
Requirement
Phase 0 runs priority spikes 1–8 and 11 before the architecture and refined estimate are committed. §28–29Not reviewed
P0-02
Requirement
Each spike states its inputs, the decision it enables, and its success criteria. §28Not reviewed
P0-03
Requirement
Missing inputs from Acolad, the Windchill owner and the publication owners are tracked as blockers with dates. App. CNot reviewed
P0-04
Requirement
Phase 0 ends with an architecture recommendation, a risk register, and a refined estimate with confidence levels. §31, §37Not reviewed
P0Q-01
Question
Who signs off the Phase 0 exit gate? §22Not reviewed
P0Q-02
Question
Is a 15-week Phase 0 (including the holiday break) acceptable, given the dependency on the Acolad transfer package? §38Not reviewed
Step 3 · Segmentation & TM leverage
SG-01
Requirement
Reproduce current Acolad segmentation closely enough to preserve TM leverage. §13Not reviewed
SG-02
Requirement
Measure boundary agreement and exact/fuzzy/no-match distributions against Acolad output on a representative corpus. §13.3Not reviewed
SG-03
Requirement
Identify the rule behind each difference and re-run after adjusting. §13.3Not reviewed
SG-04
Requirement
TM import preserves alignment, metadata and approval status; duplicates and conflicts are profiled. §14.1Not reviewed
SG-05
Requirement
TM matches are reconciled against known Acolad matches. Spike 4Not reviewed
SGQ-01
Question
What loss of exact-match leverage versus Acolad is acceptable before go-live? (Proposal: at most 1 percentage point.) §13.3Not reviewed
SGQ-02
Question
Will Acolad provide SRX rules and segmented reference output for all 38 languages, or only a sample? §13.2Not reviewed
Step 4 · AI language coverage
LC-01
Requirement
Test representative content in every language group, including RTL, CJK, Khmer, Burmese, Thai and regional locales. Spike 6Not reviewed
LC-02
Requirement
Produce a feasibility and risk profile per language. Spike 6Not reviewed
LC-03
Requirement
Verify fonts, script direction and locale formatting per language. §5Not reviewed
LC-04
Requirement
Onboard languages in waves based on the risk profile. §33Not reviewed
LCQ-01
Question
Serbian: Latin script, Cyrillic, or both? §5Not reviewed
LCQ-02
Question
Norwegian: confirm Bokmål (nb-NO). §5Not reviewed
LCQ-03
Question
Should high-risk languages (Khmer, Burmese) have a longer parallel run with Acolad? §33Not reviewed
LCQ-04
Question
Is ~2.5 hours of benchmark scoring per proofreader acceptable? §17.7Not reviewed
Step 5 · Quality validator prototype
QA-01
Requirement
Deterministic checks detect changed numbers, units, tags, placeholders and protected tokens. §17.2Not reviewed
QA-02
Requirement
Terminology validator checks required, prohibited and do-not-translate terms. §17.2Not reviewed
QA-03
Requirement
Semantic evaluator detects omissions, additions, negation, condition and sequence errors. §17.2Not reviewed
QA-04
Requirement
Findings are structured by category and severity; hard-fail rules block release regardless of any score. §17.3–17.4Not reviewed
QA-05
Requirement
The prototype is measured on seeded defects and compared with reviewer decisions. Spike 7Not reviewed
QAQ-01
Question
Which defect types must reach 100% detection before the pilot? §34Not reviewed
QAQ-02
Question
What false-positive rate is acceptable to proofreaders? (Proposal: at most 10%.) §17.7Not reviewed
Step 6 · Simulator & Windchill
IN-01
Requirement
Assess the Simulator transfer: code, build, licences, rights, runtime and security. §11.2Not reviewed
IN-02
Requirement
Recommend one option: reuse complete, reuse back end, reuse selected logic, or rewrite keeping the Windchill methods. §11.3Not reviewed
IN-03
Requirement
Demonstrate one controlled Windchill round trip with auditable states. Spike 2Not reviewed
IN-04
Requirement
Prototype one test publication per Phase 1 destination, with idempotency analysis. Spike 11Not reviewed
INQ-01
Question
Does the Acolad agreement give Schindler the right to modify and maintain the Simulator components? §11.2Not reviewed
INQ-02
Question
If Acolad’s handover is late or incomplete, do we approve the rewrite option as the fallback now? §33Not reviewed
Step 7 · Decisions & exit gate
DC-01
Requirement
Track the decisions needed before architecture, before the pilot, before reducing human review, and before Acolad exit. §34Not reviewed
DC-02
Requirement
Each decision has an owner, a due date and a linked spike. §34Not reviewed
DC-03
Requirement
The Phase 0 exit gate lists its criteria and their current status. §37Not reviewed
DCQ-01
Question
Confirm decision owners across GTL, SIS, Legal, Quality and Procurement. §22Not reviewed
DCQ-02
Question
Add “Private AI deployment model” as a decision before architecture? ReviewNot reviewed
Step 8 · Unified inbox
IB-01
Requirement
One inbox for Windchill requests and (from Phase 2) independent requests. §9.1Not reviewed
IB-02
Requirement
Views: New requests, Action required, Active operations, Human tasks, Exceptions, Completed. §9.2Not reviewed
IB-03
Requirement
Search and filter by ID, document, language, state, criticality, proofreader, destination, dates and exception type. §9.3Not reviewed
IB-04
Requirement
Actions: approve, return, hold, reject, cancel, reassign, retry, restart from an authorized stage, open history, export. §9.4Not reviewed
IB-05
Requirement
Any view exports to XLSX or CSV, including its filters and a timestamp. §9.5Not reviewed
IBQ-01
Question
Who besides GTL Operations needs inbox access (for example content owners, read-only)? §22Not reviewed
IBQ-02
Question
Should the default view show requests, or individual language jobs? Not reviewed
Step 9 · Request & pre-flight
RQ-01
Requirement
A Windchill request carries source identity, revision, target languages, business metadata and initial status. §8.2Not reviewed
RQ-02
Requirement
Windchill decides which topic versions need translation per language; topics already translated or in translation are excluded. §7.3Not reviewed
RQ-03
Requirement
Pre-flight checks state, current revision, languages, duplicates, topics in translation, metadata, package generation, destination and interface availability. §8.3Not reviewed
RQ-04
Requirement
Pre-flight ends in one of: ready for GTL validation, action required, technical exception, duplicate suspected, source obsolete. §8.3Not reviewed
RQ-05
Requirement
Before approval, show a forecast of TM leverage, proofreader effort and completion date. §33Not reviewed
RQQ-01
Question
May criticality be inferred automatically (for example from hazard statements) and only confirmed by GTL? §8.4Not reviewed
RQQ-02
Question
Should an unavailable proofreader block approval, or only raise a warning? §24Not reviewed
Step 10 · GTL validation
AP-01
Requirement
GTL can approve, approve with controlled changes, return for information, hold, reject or cancel. §8.4Not reviewed
AP-02
Requirement
Controlled changes: target languages, priority, criticality, translation profile, proofreader policy, output profile, destination. §8.4Not reviewed
AP-03
Requirement
The gate stays lightweight: the system pre-fills recommendations and highlights only exceptions. §8.4Not reviewed
AP-04
Requirement
Every decision is recorded with user, time and reason. §26.1Not reviewed
APQ-01
Question
Which request profiles could later be approved automatically (straight-through)? §8.4Not reviewed
APQ-02
Question
Is a reason mandatory for every change, or only for reject and cancel? Not reviewed
APQ-03
Question
For High criticality, should exact matches in hazard statements also be shown to the proofreader for confirmation? §8.14Not reviewed
Step 11 · Language jobs
MN-01
Requirement
One language job per source revision, target locale, output profile and destination. §7.2Not reviewed
MN-02
Requirement
Language job states are visible for all 38 languages. §23.2Not reviewed
MN-03
Requirement
Completed and failed language jobs are tracked separately; one language’s problem does not block the others. §24Not reviewed
MN-04
Requirement
Topic-level reuse and segment-level reuse are measured independently. §7.3Not reviewed
MN-05
Requirement
Human waiting time is measured separately from system processing time. §25.3Not reviewed
MNQ-01
Question
May each language publish as soon as it is ready, or must all 38 publish together? §24Not reviewed
MNQ-02
Question
What turnaround per language job should trigger an escalation? §26.2Not reviewed
Step 12 · Segments & TM matching
JB-01
Requirement
Each unit is classified: approved exact, exact with context warning, exact conflict, fuzzy (high/medium/low), no match, TM error. §14.2Not reviewed
JB-02
Requirement
Approved exact matches are reused without AI but still pass terminology, protected-content and integrity checks. §8.8Not reviewed
JB-03
Requirement
Fuzzy matches are evidence for the AI, never approved output. §8.8Not reviewed
JB-04
Requirement
The AI request includes terminology, protected tokens, context and constraints; model, prompt and terminology versions are recorded. §8.9Not reviewed
JB-05
Requirement
Segment-level validation runs before reconstruction and produces structured findings. §8.10Not reviewed
JBQ-01
Question
Which fuzzy thresholds separate high, medium and low? (Proposal: 95 / 85 / 75%.) §14.2Not reviewed
JBQ-02
Question
When two approved exact matches conflict, may the most recent entry from the same domain win automatically? §14.3Not reviewed
Step 13 · AI correction loop
LP-01
Requirement
Correctable findings go back to the AI with explicit instructions; all attempts and findings are kept. §8.11Not reviewed
LP-02
Requirement
The loop stops on: pass, hard fail, no improvement, oscillation, maximum attempts, source ambiguity, missing terminology. §16.3Not reviewed
LP-03
Requirement
Unresolved units are marked for mandatory human correction. §8.11Not reviewed
LP-04
Requirement
The maximum number of attempts is configurable. §2.2Not reviewed
LPQ-01
Question
May the loop fix a hard-fail numeric deviation itself, or must every hard fail go straight to a person? §17.4Not reviewed
LPQ-02
Question
Maximum number of correction attempts in Phase 1? (Proposal: 2.) §2.3Not reviewed
Step 14 · Proofreader workbench
WB-01
Requirement
Edit per segment, review in the reconstructed document, approve per language job. §18.1Not reviewed
WB-02
Requirement
Side-by-side source and target with a synchronised preview, terminology, TM provenance and findings. §18.2Not reviewed
WB-03
Requirement
Filters for content that needs review; exact matches are shown as context only. §8.14Not reviewed
WB-04
Requirement
Accept, edit, reject, comment, escalate, raise a source query, propose a term. §8.14Not reviewed
WB-05
Requirement
Keyboard-first review, autosave, version history, and an audit of every change. §18.2Not reviewed
WB-06
Requirement
All target scripts, including right-to-left, display and edit correctly. §25.6Not reviewed
WBQ-01
Question
Must proofreaders confirm every unit, or may they approve once all findings are resolved? §8.14Not reviewed
WBQ-02
Question
Should reviewers tag every edit with an error category, to build the data for later quality calibration? §17.7Not reviewed
WBQ-03
Question
Do any proofreaders need to export to their own CAT tool (XLIFF) instead of using the workbench? Not reviewed
Step 15 · Source query
SQ-01
Requirement
A proofreader can raise a query on ambiguous English; the affected unit is paused. §18.3Not reviewed
SQ-02
Requirement
The content owner either clarifies or creates a new source revision. §18.3Not reviewed
SQ-03
Requirement
A new source revision invalidates the affected target work and restarts matching, translation and review. §18.3Not reviewed
SQQ-01
Question
What response time should content owners commit to? Not reviewed
SQQ-02
Question
Should a clarification go to all 38 proofreaders automatically? Not reviewed
Step 16 · Final validation
FV-01
Requirement
After human changes, deterministic and terminology checks run again, edited units are reinserted, and the output is rebuilt. §8.15Not reviewed
FV-02
Requirement
Integrity: all units present, no duplicates, correct order, protected content unchanged, tags valid, import succeeds. §8.13Not reviewed
FV-03
Requirement
Publication stays blocked until all mandatory checks pass. §8.15Not reviewed
FV-04
Requirement
Only fully approved content updates the TM, with provenance. §8.16Not reviewed
FV-05
Requirement
Terminology proposals go to the terminology workflow, not straight into the baseline. §8.16Not reviewed
FVQ-01
Question
Is a GTL sign-off needed after the proofreader approves, or is the proofreader’s approval final? §8.15Not reviewed
Step 17 · Return & publication
PB-01
Requirement
Windchill return: import XLIFF, link target topics to the source revision, publishPDFDraft, validate, publishDoc, confirm target state. §20.2Not reviewed
PB-02
Requirement
SAP/PNC: language-specific identifier, metadata, link to the English master, release status. §20.3Not reviewed
PB-03
Requirement
Each publication records request, job, source revision, locale, destination record, checksum, connector version and result. §20.1Not reviewed
PB-04
Requirement
Publication is idempotent; a failed publication never marks the job as complete. §8.17Not reviewed
PBQ-01
Question
What are the language-specific ID rules in SAP/PNC (for example Serbian Latin vs Cyrillic)? §2.3Not reviewed
PBQ-02
Question
Wording and placement of the visible translation label in the PDF? §21Not reviewed
Step 18 · Audit & evidence
AU-01
Requirement
Mandatory audit data is kept for every request and language job. §26.1Not reviewed
AU-02
Requirement
Minimum evidence is kept for every translation unit. App. BNot reviewed
AU-03
Requirement
Traceability is stored as metadata, not only as visible text in the document. §21Not reviewed
AUQ-01
Question
How long must audit evidence and AI attempts be retained? §25.1Not reviewed
AUQ-02
Question
Who gets read-only auditor access? §22Not reviewed
Step 19 · Exceptions
EX-01
Requirement
Each exception type follows the behaviour in the catalogue: queue and retry, block, or route to an owner. §24Not reviewed
EX-02
Requirement
Safe retry and restart from an authorized stage. §9.4Not reviewed
EX-03
Requirement
Exceptions show age, owner and SLA status. §26.2Not reviewed
EXQ-01
Question
Who owns each exception type (GTL, SIS support, terminology owner, publication owner)? §22Not reviewed
EXQ-02
Question
Escalation path when no proofreader is available for a language? §24Not reviewed
Step 20 · Translation memory
TM-01
Requirement
Import all 38 TMs, keeping alignment, metadata and approval status. §14.1Not reviewed
TM-02
Requirement
Detect duplicates and conflicts, and assign trust states. §14.1Not reviewed
TM-03
Requirement
Update only with approved content: never draft AI output, failed attempts, unresolved findings or obsolete sources. §14.4Not reviewed
TM-04
Requirement
Export in a standard format, back up and restore, keep history and provenance. §14.1Not reviewed
TMQ-01
Question
Who resolves TM conflicts: the proofreader, or a TM owner in GTL? Not reviewed
TMQ-02
Question
Should low-trust legacy entries be quarantined, or demoted to fuzzy evidence? §33Not reviewed
Step 21 · Terminology
TE-01
Requirement
A periodic TermWeb export is imported as a read-only baseline. §15.2Not reviewed
TE-02
Requirement
Controlled local additions and overrides, always distinguishable from imported records. §15.3Not reviewed
TE-03
Requirement
A re-import never silently overwrites approved local changes; conflicts go to the terminology owner. §15.4Not reviewed
TE-04
Requirement
Roles: viewer, contributor, approver, administrator, proofreader with proposal rights. §15.5Not reviewed
TE-05
Requirement
Only terms relevant to a unit are sent to the AI model. §8.7Not reviewed
TEQ-01
Question
How often should TermWeb be imported (weekly, monthly)? §15.1Not reviewed
TEQ-02
Question
Should approved local terms be fed back into TermWeb, and by whom? §4.4Not reviewed
TEQ-03
Question
Who are the terminology approvers per language? §34Not reviewed
Step 22 · Reports & KPIs
RP-01
Requirement
Business KPIs: cost avoided, scope replaced, reuse rates, AI and human correction, turnaround, findings, publication success. §3.3Not reviewed
RP-02
Requirement
Operational, quality and business reports. §26.2–26.4Not reviewed
RP-03
Requirement
All detailed reports export to XLSX or CSV. §26Not reviewed
RPQ-01
Question
Which KPIs go into the CEO dashboard? §32Not reviewed
RPQ-02
Question
Turnaround and cost baseline: from Acolad invoices, or measured during Phase 0? §3.3Not reviewed
Step 23 · Policies & versions
CF-01
Requirement
Model, prompt, terminology, rule and parser versions are registered and recorded with every attempt. §8.9, §25.5Not reviewed
CF-02
Requirement
Initial review policy: new, fuzzy and AI-corrected content is always reviewed; exact matches are reused with checks. §17.8Not reviewed
CF-03
Requirement
Correction loop limits are configurable. §8.11Not reviewed
CF-04
Requirement
Hard-fail rules block release regardless of any score. §17.4Not reviewed
CF-05
Requirement
Every production model or rule change runs the approved regression suite. §30Not reviewed
CF-06
Requirement
No silent fallback to an unapproved model: if the AI service is down, work queues. §16.1, §24Not reviewed
CFQ-01
Question
Who approves model and prompt changes (AI model owner, GTL, both)? §22Not reviewed
CFQ-02
Question
Which content categories always need full review, regardless of score? §8.14Not reviewed

Every screen lists its items in the Validate panel. Back to the start