CertOf Enterprise Standard
CertOf Visual Equivalence Evaluation Standard for Translation Outputs
| Item | Content |
|---|---|
| Document number | CERTOF-STD-VEE-001-2026 |
| Version | v0.11 first public English edition |
| Publisher | CertOf |
| Authors | Xener and Xi |
| Publication date | 2026-05-28 |
| Effective date | 2026-05-28 |
| Reference verification date | 2026-05-28 |
| Purpose | To rate the textual, structural, visual and informational correspondence between a source document and a translation output. |
| Important notice | This document evaluates translation output. It does not evaluate translation service processes, production methods, translator qualifications, authenticity of source documents or legal effect. |
| Status | First public English edition. |
This document is drafted as a CertOf enterprise standard for global use. It defines a general evaluation method for visual equivalence between a source document and a translation output.
This document uses a common standards-document structure to improve readability, review consistency and external referenceability. Unless CertOf later publishes or declares otherwise, this document does not claim compliance with any national or regional filing rule, enterprise-standard self-declaration system or mandatory regulatory requirement.
This document does not specify commercial translation-service processes, order management, target-grade commitments, delivery time, pricing, internal production methods, translation tools, layout methods or translator qualifications. It specifies only evaluation objects, evidence scope, format and carrier records, quality classes, decision logic and minimum report fields.
The provisions on object retention, visual restoration, translation marking, table mapping and non-text elements are inspection and decision logic. They are not production instructions. Evaluators shall judge an already formed translation output and shall not interpret this document as a set of required production steps.
Class A, Class B, Class C, Class D and Class U describe actual evaluation results for an already formed translation output. An evaluation result shall not be understood as a guarantee of provider qualification, commercial commitment, source-document authenticity, legal effect or fitness for a particular use.
1 Scope and Boundaries
2 Normative References
3 Terms and Definitions
4 General Principles
5 Evaluation Evidence and Scope of Conclusion
6 Assessment Carrier, Input Format, Output Format and Editable Status
6.1 Output Editable Status
7 Class X Result Model and Decision Rules
7.1 Core Dimensions
7.2 Lowest-Dimension Principle
7.3 Class U Override
8 Text Quality (TQ)
8.1 CertOf Default TQ Quantitative Model
8.2 CertOf Default TQ Class Mapping
9 Visual Equivalence (VE)
9.1 VE-A Trusted Visual Restoration Logic
10 Information Mapping (IM)
10.1 Non-Text Visual Object Check Logic
11 Multilingual Synchronization Evaluation (Optional)
12 AI and Automation Output Risk Notice
13 Result Expression and Minimum Report Fields
Annex A (Normative): Class U Triggers
Annex B (Normative): CertOf Default STU Counting Configuration Requirements
Annex C (Informative): Minimum Closed-Loop Example
Annex D (Informative): Quick Evaluation Checklist
Annex E (Informative): Informative Reference Verification Record
This document specifies methods for evaluating text quality, structural correspondence, visual equivalence, information mapping, multilingual synchronization and related status information between a source document and a translation output.
This document applies to evaluation of the following objects:
This document does not apply to evaluation of a translation-service provider's organizational capability, personnel qualifications, pricing, delivery time, customer service or internal process. Such matters may be specified by other service standards, contracts or management rules.
Video, audio, real-time subtitles, dubbing and other dynamic media may use the text-quality, information-mapping and synchronization principles in this document for limited evaluation. This document is not a complete evaluation standard for dynamic-media translation.
The following referenced document is indispensable for the application of this document. For dated references, only the cited edition applies. For undated references, the latest edition of the referenced document applies.
| Document | Title | Use in this document |
|---|---|---|
| ISO 5060:2024 | Translation services - Evaluation of translation output - General guidance | Used as the analytic evaluation framework for Text Quality (TQ). Based on the ISO 5060:2024 approach to error types, severity, penalty points or weights, error scores and quality-rating systems, this document defines the CertOf default TQ class mapping. |
Except for ISO 5060:2024, other international standards mentioned in this project are informative references only and do not constitute indispensable normative provisions of this document.
The reference verification date of this document is 2026-05-28. The official ISO page for ISO 5060:2024 identifies the document as Published, Edition 1, with publication date 2024-02.
| Term | Definition |
|---|---|
| source document | The original document, image, web page, file, PDF, slide deck, form or other content carrier used as the basis for translation. |
| translation output | An output that presents the source-document content in another language. It may be produced by an original issuer, authorized issuer, third-party translator, machine translation system, AI-assisted system or mixed workflow. |
| translated copy | A translation output produced by a party other than the original publisher or issuing authority. This document evaluates only its textual, structural, visual and informational correspondence with the source document. |
| visual equivalence | The degree to which the source document and translation output are comparable or approximately consistent in page appearance, layout, major elements, reading path, information relationships and functional perception. |
| verified visual basis | A visual basis formed from the visible information of the source document and from verifiable formal layout materials, such as official public samples, formal templates or source files from the original publisher. |
| non-text visual object | A photo, signature, seal, badge, QR code, barcode, mark, background pattern, watermark, graphic, chart or other non-body-text object that carries visual or factual relationships. |
| assessment carrier | The material form actually used by the evaluator for inspection, such as an official PDF, rendered image, web preview, screenshot or customer-provided image. |
| delivery format | The file form actually received by the customer or declared as delivered, such as PDF, DOCX, PPTX, web page, image or paper copy. |
| visible-evidence conclusion | A limited evaluation conclusion formed from visible pages, samples or previews when the evaluator does not have the complete official file or full-text TQ evidence. A visible-evidence conclusion shall not be treated as an official Class X result. |
| NR, no rating | No official class is issued. NR is not a Class X quality class and does not mean conforming or non-conforming. It means that no official Class X was formed because of evidence scope, evaluation scope or engagement limitations. |
| Class X | The overall evaluation class under this document. X is A, B, C, D or U. A is the highest class, D is the lowest conforming class and U is non-conforming. |
| TQ / VE / IM | Text Quality, Visual Equivalence and Information Mapping. |
The evaluation object is the result relationship between the source document and the translation output, not the production process of the service provider. This document is designed for global use and uses only the international standards expressly listed in this document as normative references.
Class X is an actual evaluation result. This document does not define a management concept of a target class or achieved target class. Text Quality, Visual Equivalence and Information Mapping are the core quality dimensions of Class X.
Inspection items in this document are used only to judge an already formed translation output. They do not specify production steps, tool choices, layout methods or visual-asset generation methods. The translation output may be produced by an original issuer, third-party translator, machine translation system, AI-assisted system or mixed workflow. This document does not require disclosure of the translation process or tools.
An official Class X evaluation shall cover the full source document and the full translation output. Sampling shall not replace full-document evaluation. This document adopts the text-quality evaluation framework of ISO 5060:2024 but does not adopt sampling as a substitute for full-document evaluation.
If the evaluator has only a web preview image, public sample, compressed screenshot or partial material, the evaluator shall issue a visible-evidence conclusion or NR and shall not express the result as an official full-document Class X evaluation.
A translated copy shall be distinguished from the original by wording such as "translation", "translated copy" or an equivalent marking. For translation-output evaluation, provided that factual misunderstanding or forgery risk is not created, closer correspondence with the visual and informational relationships of the source document supports a higher VE class.
Any error that seriously affects core facts, key fields or the source-target correspondence may trigger Class U. Class U is the only non-conforming status in this document. Insufficient evidence caused by missing evaluation material is not itself Class U, but it limits or prevents issuance of an official Class X result.
Normative terms in this document shall be interpreted as follows: "shall" and "shall not" indicate requirements; "should" and "should not" indicate recommendations; "may" indicates permission; "can" indicates capability, possibility or an explanatory statement and does not itself create a requirement.
Evaluation evidence describes what material the evaluator actually inspected and the scope to which the conclusion may apply. Evidence scope does not directly rate translation quality, but it determines whether an official Class X can be issued.
| Code | Status | Decision basis | Permitted conclusion |
|---|---|---|---|
| ES-FULL | Full official evaluation | The evaluator has the official source document and official translation output and can check the full text, all pages, all key fields and delivery format. | Official Class X may be issued. |
| ES-VISIBLE | Visible-page evaluation | The material is sufficient to inspect VE and IM on visible pages but is not sufficient to confirm full-text TQ, all pages or the official delivered file. | A visible-evidence conclusion may be issued. Class X shall be NR and shall not be expressed as an official Class X. |
| ES-SAMPLE | Public sample or preview evaluation | The material comes from a web sample, compressed image, public example, screenshot or partial display. | A candidate class or risk observation may be issued. Class X shall be NR and shall not be expressed as an official Class X. |
| ES-INSUFFICIENT | Insufficient evidence | The evaluator lacks sufficient material to establish a basic source-target relationship or to judge the main quality dimensions. | Class X shall be NR. The report shall record missing evidence and the reason why an official class cannot be issued. |
A certificate of translation accuracy, translator declaration or delivery-package statement may be recorded as an evidence clue and metadata. It shall not automatically replace full-text TQ evaluation under ISO 5060. Whether a certificate page is included in VE/IM evaluation depends on whether the evaluation object is a single translation output or the complete delivery package.
If a certificate page or declaration page is included in evidence records, the evaluator should check whether language direction, file name or description, page count, date, translator or provider declaration and coverage are consistent with the actual source document and translation output. If certificate coverage is inconsistent, the limitation shall be recorded. If it creates a key factual error or misleading delivery package, it may be handled under TQ, IM or Class U rules.
Insufficient evidence should not be mechanically rated as Class U. Class U shall be assigned only when the evaluator has the official source document and official translation output and the evaluated object itself remains unopenable, unviewable, contains key factual errors, lacks an establishable source-target factual correspondence or meets another non-conformity trigger.
Input format, output format and assessment carrier describe the form of the object inspected by the evaluator. They do not directly indicate quality, but they affect evidence scope, VE-A trusted visual restoration, file usability and report traceability.
| Code | Name | Description |
|---|---|---|
| IF-NATIVE-DOC | Native editable document | DOCX, PPTX, XLSX, HTML, design source file or similar file in which text, structure and objects can be directly read or edited. |
| IF-PDF-NATIVE | Native PDF | PDF containing extractable text, objects or structural tags. |
| IF-PDF-SCAN | Scanned PDF | PDF mainly composed of scanned images; text is not directly extractable or is obtained only by OCR. |
| IF-IMAGE-SCAN | Scanned image | Image file produced by a scanner or scanning application. |
| IF-IMAGE-PHOTO | Photographed image | Image captured by a camera or mobile device, potentially containing perspective distortion, uneven lighting, blur or occlusion. |
| IF-WEB | Web page or web application | HTML page, web application interface or web-rendered output. |
| IF-OTHER | Other input | Other source form. The report shall describe the actual form and any evaluation limitations. |
Output format categories describe the file form actually delivered to or declared for the customer.
| Code | Name | Description |
|---|---|---|
| OF-PDF | PDF output | PDF used as the main delivery format. |
| OF-IMAGE | Image output | PNG, JPG, WebP, TIFF or similar image output. |
| OF-NATIVE-DOC | Native editable output | DOCX, PPTX, XLSX, HTML, design source file or similar editable output. |
| OF-WEB | Web output | Web page, web application interface or hosted translation output. |
| OF-PAPER | Paper output | Printed or physically delivered translation output. |
| OF-MIXED | Mixed output | A delivery package containing multiple output formats. |
| OF-OTHER | Other output | Other delivery form. The report shall describe the actual form and any evaluation limitations. |
Editable status describes whether the customer can reasonably edit text, objects or layout in the delivered translation output. It does not directly determine quality class, but it shall be recorded when it affects usability, evidence scope or customer interpretation.
| Code | Status | Description |
|---|---|---|
| OE-EDITABLE | Editable | Text, objects or layout can be reasonably edited in ordinary professional software. |
| OE-PARTIAL | Partly editable | Some text, objects or layout can be edited, but important portions are flattened, embedded or otherwise restricted. |
| OE-FLAT | Flattened | The output is mainly a flattened page image or non-editable rendering. |
| OE-LOCKED | Locked or restricted | The output contains permissions, passwords or technical restrictions that prevent ordinary editing or inspection. |
| OE-N/A | Not applicable | Editable status is not relevant to the evaluated delivery format or engagement scope. |
Class X is the overall result of this document. It is determined by the core dimensions TQ, VE and IM, subject to evidence-scope limitations and Class U triggers.
| Class | Meaning | General description |
|---|---|---|
| Class A | Excellent visual equivalence | The output reaches high text quality, highly equivalent visual presentation and complete information mapping. It may be based on verified restoration or third-party translation if the result is equivalent to the source and properly marked as a translation. |
| Class B | Good visual equivalence | The output is reliable for ordinary comparison and use. Minor differences may exist but do not affect key facts, reading path or main information relationships. |
| Class C | Basic visual correspondence | The output preserves the main content and information relationships, but visual or structural differences are noticeable and may reduce comparison convenience. |
| Class D | Minimum conforming correspondence | The output remains usable for limited understanding or reference, but text, visual or mapping quality is weak. It does not trigger Class U. |
| Class U | Non-conforming | The output fails because of a Class U trigger. Class U is the only non-conforming status under this document. |
The core dimensions shall be recorded as TQ, VE and IM. Each dimension shall be assigned A, B, C, D or U when official evidence is sufficient.
| Dimension result | Class X effect | Notes |
|---|---|---|
| TQ-A / VE-A / IM-A | Supports Class A | Class A requires all core dimensions to reach A and no Class U trigger. |
| Any core dimension B and no lower result | Highest possible Class B | The overall class shall not exceed the lowest core-dimension class. |
| Any core dimension C and no lower result | Highest possible Class C | Class C remains conforming if no Class U trigger exists. |
| Any core dimension D and no U result | Highest possible Class D | Class D is the lowest conforming class. |
| Any core dimension U or any Class U trigger | Class U | Class U overrides A/B/C/D mapping. |
| Evidence insufficient for official evaluation | NR | NR is not a quality class and is not Class U by itself. |
When official evidence is sufficient, the overall Class X shall not be higher than the lowest result among TQ, VE and IM. If any dimension is U, or if a separate Class U trigger applies, the overall result shall be Class U.
Contextual status codes such as evidence scope, input format, output format, editable status and synchronization status shall be reported separately. They shall not raise a Class X result above the core-dimension result.
Class U is a non-conforming result. It may be triggered by severe text errors, inability to establish a source-target visual or structural correspondence, unviewable files, material factual mismatch, missing key information, misleading visual object handling or other conditions listed in Annex A.
There is no class lower than VE-D, TQ-D or IM-D. If the evaluator cannot establish the visual, structural or factual correspondence required for a D-level result, the result shall be U rather than a lower letter class.
Text Quality evaluates whether the target-language text is accurate, complete, understandable, terminologically consistent and fit for the corresponding class of use. TQ uses the analytic evaluation approach described in ISO 5060:2024: error types, severity, penalty points or weights, error scores and quality-rating systems are used to form the evaluation conclusion.
ISO 5060:2024 shall not be understood as already defining globally fixed thresholds for TQ-A, TQ-B, TQ-C, TQ-D and TQ-U under this document. This document defines the CertOf default TQ class mapping as an enterprise-standard configuration. If the client or evaluation project uses another recorded ISO 5060-compatible configuration, the evaluation report shall disclose it and explain its mapping to TQ-A/B/C/D/U.
TQ shall be evaluated on the full translation output when official Class X is issued. Sampling shall not be used as a substitute for full-document evaluation under this document.
The evaluation report shall record the ISO 5060:2024-compatible text-quality configuration used, including error categories, severity, penalty points or weights, normalization denominator, quality-rating thresholds, whether a client-specific configuration exists and how TQ-A/B/C/D/U map to that configuration. If the evaluation configuration is not recorded, an official TQ class should not be issued.
If no project-specific configuration is declared, the CertOf default TQ quantitative model may be used. This model is an ISO 5060-compatible configuration and is not a fixed threshold provided by ISO 5060.
Formula 1: TQ normalized error score. S = P / Ne * 1000. P is the total evaluation penalty score and Ne is the effective normalization denominator. The denominator of the CertOf default TQ profile uses only standard text units (STU). Character count, line count or information-unit count shall not be directly substituted into this default threshold model.
Formula 2: Effective normalization denominator. Ne = max(STU, 1000). To avoid score inflation in short certificates, short attestations or fixed-field documents caused by a very small number of minor errors, when STU is less than 1000 the actual STU and Ne shall both be recorded. Key-field errors, Critical errors and IM-U risks are not offset by short-text calibration.
Formula 3: Total TQ penalty score. P = sum(w * m). Here w is the error-type weight and m is the severity multiplier. By default, the base weight w for each error type is 1. If project-specific error-type weights are used, they shall be recorded in the report.
| Severity | Default multiplier m | Description |
|---|---|---|
| Minor | 1 | A localized error that does not affect key facts, main meaning or main usability. |
| Major | 5 | An error that affects meaning, completeness, terminology, readability or important information, but does not by itself trigger Class U. |
| Critical | U trigger | An error that seriously affects core facts, legal or factual relationships, identity, amount, date, qualification, right or obligation, or source-target correspondence. A Critical error may directly trigger TQ-U and Class U. |
The STU counting method shall be declared in the report. When the CertOf default TQ profile is used, the STU counting configuration requirements in Annex B shall also be used. Languages that are primarily segmented by spaces may be counted by word units. Languages such as Chinese, Japanese and Thai that are not stably segmented by spaces shall use a reproducible and version-locked tokenizer, rule set or manual word-unit method, and shall record tool name, rule name, version or manual segmentation principles. If a stable and reproducible STU method cannot be formed, a project-specific TQ profile shall be used and the CertOf default thresholds shall not be directly applied.
The following mapping applies only when the CertOf default TQ quantitative model is used and the evaluator has official full-text evidence.
| TQ class | Default quantitative condition | Additional condition |
|---|---|---|
| TQ-A | S <= 2 | No Critical error, no key-field error and no material omission. |
| TQ-B | S <= 5 | No Critical error. No error may materially change a key fact or the main meaning. |
| TQ-C | S <= 10 | No Critical error. The output remains understandable and usable for general reference. |
| TQ-D | S <= 20 | No Critical error. The output remains minimally usable and does not meet any TQ-U trigger. |
| TQ-U | Critical error or S > 20 | Any error that seriously affects core facts, key fields or source-target correspondence may trigger TQ-U regardless of S. |
If the evaluator uses a project-specific ISO 5060-compatible configuration, the report shall record the configuration and its mapping to TQ-A/B/C/D/U. The report shall not imply that the project-specific thresholds are fixed ISO 5060 thresholds.
Visual Equivalence evaluates whether the translation output preserves, restores or reasonably approximates the source document's page appearance, layout, element relationships, visual hierarchy and reading path.
VE evaluates the result. It does not prescribe how to perform layout, restoration or visual asset preparation. A high VE result may be produced by the original issuer, an authorized party or a third party if the output is based on verifiable visual evidence and is clearly marked as a translation output where applicable.
| VE class | Decision logic |
|---|---|
| VE-A | The translation output is visually equivalent to the source document or to a trusted restored version of the source document. Page layout, major objects, hierarchy, spacing, alignment, reading path and functional perception are highly consistent. If the source is a photographed or distorted image, VE-A may be based on restoration of photographic problems such as perspective distortion, skew, uneven lighting or lens effects, provided that the restoration is based on visible source information and verified formal visual basis. |
| VE-B | The output preserves the main layout and visual relationships. Minor differences exist but do not affect comparison, reading path, key information or main functional perception. |
| VE-C | The output preserves the main page structure and major objects, but visual differences are obvious and reduce comparison convenience. |
| VE-D | The output has only minimum visual correspondence. It remains possible to identify the main content relationship, but layout, hierarchy or object placement is weak. |
| VE-U | The evaluator cannot establish a visual or structural correspondence sufficient for VE-D, or the output has severe visual object errors, misleading substitution, missing key visual elements or unviewable pages. |
Trusted visual restoration is part of VE-A decision logic. It applies when the source material is visually degraded by capture conditions, such as a photographed document, skewed scan, perspective distortion, uneven lighting or blur, and the translation output restores the document to a visual effect equivalent to the original formal document.
| Check item | Decision requirement |
|---|---|
| Source basis | The restoration shall be based on visible source information and verifiable formal visual basis, such as official public samples, formal templates, original source files or other reliable layout evidence. |
| Distortion correction | Correction of perspective distortion, skew, lighting imbalance, crop and capture artifacts may support VE-A if it restores the formal document effect instead of inventing content. |
| No unverified completion | The restoration shall not add unverified facts, signatures, seals, marks, photographs, numbers, dates or other content that is not supported by the source or verified visual basis. |
| Translation marking | If the output is a translated copy rather than an original reissued document, it shall be distinguishable as a translation output by marking or equivalent means. |
| Report record | The report shall record the source condition, restoration basis and any visual assumptions or limitations material to the VE-A decision. |
If restoration changes facts, adds unverifiable objects, removes key visible information or makes the output appear to be an original document when it is only a translated copy, the result shall not be VE-A and may trigger Class U depending on severity.
Information Mapping evaluates whether source-document information units, labels, values, relationships, order and non-text visual objects are correctly mapped to the translation output.
IM focuses on the checkable correspondence between source and target information. It does not require the evaluator to know the production method used to create the output.
| IM class | Decision logic |
|---|---|
| IM-A | All key information units and relationships are mapped completely and correctly. Order, grouping, labels, values, page relationships and non-text visual objects support reliable comparison. |
| IM-B | Main information units and relationships are mapped correctly. Minor omissions or differences exist but do not affect key facts or main use. |
| IM-C | Main information is present, but mapping differences, weak grouping or partial omissions reduce comparison convenience. |
| IM-D | Minimum mapping exists. The main source-target relationship can still be identified, but mapping quality is weak and suitable only for limited reference. |
| IM-U | Key information is missing, mismapped, misleadingly grouped, assigned to the wrong entity or otherwise impossible to establish at the minimum level required for IM-D. |
For forms, statements, transcripts, legal documents, registration documents, contract tables, line-item lists and multi-page documents, the evaluator shall check not only isolated text but also information relationships.
| Relationship type | Check logic |
|---|---|
| Label-value relationship | Labels and values shall remain correctly associated. A translated value shall not be assigned to the wrong label. |
| Row-column relationship | Rows, columns, headers, subtotals and totals shall preserve their source relationships. |
| Entity relationship | Names, identification numbers, organizations, roles, signatures, seals and dates shall remain associated with the correct entity. |
| Sequence relationship | Page order, paragraph order, item order and procedural order shall not mislead the reader. |
| Reference relationship | Footnotes, cross-references, QR codes, barcodes, certificate numbers and related statements shall remain correctly connected with the referenced information. |
Non-text visual objects shall be checked as evidence-bearing objects when they carry factual, identity, verification, structural or visual-comparison significance. This section specifies inspection logic only and does not prescribe how such objects are produced.
| Object type | Inspection logic |
|---|---|
| Photographs and portraits | Check whether the object is retained, omitted, redacted or replaced and whether the treatment affects identity or source-target correspondence. |
| Signatures and seals | Check whether visible signatures, seals or stamps are represented in a way that supports comparison without misleading the reader into treating the translated copy as an original issuance. |
| QR codes and barcodes | Check whether the object is retained, translated, replaced, disabled or described, and whether the report records limitations when machine verification is not assessed. |
| Logos, badges and official marks | Check whether the object is present, omitted, replaced or distorted and whether the treatment affects identification, authority perception or visual comparison. |
| Backgrounds, patterns and watermarks | Check whether the treatment affects readability, anti-confusion marking, document comparison or perceived document status. |
| Charts, diagrams and graphics | Check whether labels, legends, values, scale, direction and data relationships are correctly mapped. |
Non-text visual object checks shall produce an object register when such objects are material to the evaluation.
| Field | Requirement |
|---|---|
| Object identifier | Record page, region, object type or other stable identifier. |
| Source status | Record whether the object is visible, partial, unclear, redacted or absent in the source material. |
| Output treatment | Record whether the object is retained, omitted, translated, described, redacted, replaced or otherwise changed. |
| Evaluation effect | Record whether the treatment affects TQ, VE, IM or Class U. |
| Limitation | Record any limitation, such as no machine verification of a QR code or lack of authority to verify a seal. |
This section applies only when the evaluated object is a multilingual content asset with version-update relationships, such as a website, application interface, manual, product documentation set or continuously maintained multilingual publication.
| Code | Status | Description |
|---|---|---|
| SYNC-A | Fully synchronized | Source and target versions, pages, modules, links, visible content and key updates are synchronized. |
| SYNC-B | Mostly synchronized | Minor delay or minor difference exists but does not affect key information or main use. |
| SYNC-C | Partly synchronized | Some modules, pages or update relationships are inconsistent and reduce usability. |
| SYNC-D | Weakly synchronized | Minimum synchronization exists, but significant update gaps or structural differences remain. |
| SYNC-U | Synchronization failure | Version mismatch, missing target content or wrong-source mapping makes the multilingual asset materially misleading or unusable. |
| SYNC-N/A | Not applicable | The evaluated object is a static one-time document or synchronization is outside the engagement scope. |
SYNC status shall not raise Class X above the TQ/VE/IM result. If synchronization failure creates key factual mismatch or missing content, it may affect TQ, IM or Class U.
This document evaluates final translation output and does not require disclosure of whether AI, machine translation, OCR, layout automation or other tools were used.
If the final output contains hallucinated content, unsupported completion, mistranslated key facts, invented visual objects, incorrect entity mapping, unverified restoration, broken reading order or other result defects, the evaluator shall assess the defect under TQ, VE, IM or Class U regardless of production method.
Use of AI or automation is not itself a quality defect. Conversely, absence of AI or automation does not itself establish conformity.
An official evaluation report shall express the overall result as Class X and shall separately record TQ, VE, IM and applicable status codes. If official evidence is insufficient, the report shall use NR and describe the evidence limitation.
The report shall include at least the following fields.
| Field | Minimum requirement |
|---|---|
| Evaluated object | Identify the source document, translation output, language direction, file names or stable identifiers, page range and delivery format. |
| Evidence scope | Record ES-FULL, ES-VISIBLE, ES-SAMPLE or ES-INSUFFICIENT and describe any limitation. |
| Input, output and carrier | Record input format, output format, assessment carrier and editable status. |
| TQ result | Record the ISO 5060:2024-compatible evaluation configuration, CertOf default TQ profile or project-specific profile, error penalties, actual STU, effective denominator Ne, normalized score S and TQ class. |
| VE result | Record VE class, visual basis, major layout findings, restoration basis if VE-A uses trusted restoration, and any visual limitation. |
| IM result | Record IM class, key information mapping findings, non-text object treatment and any mapping limitation. |
| Class U check | Record whether any Class U trigger exists and the reason if Class U is assigned. |
| Overall result | Record Class A, B, C, D, U or NR, applying the lowest-dimension principle and evidence limitations. |
| Date and evaluator | Record evaluation date, evaluator or evaluating organization and any relevant version information. |
When a report is based only on visible evidence, sample material or public preview material, it shall state that the conclusion is limited and shall not present the result as an official full-document Class X.
Class U shall be assigned when any of the following conditions applies to the evaluated object and official evidence is sufficient to make the determination.
Evidence insufficiency alone is not Class U. If the evaluator lacks sufficient material to determine conformity, the result shall be NR unless the available official evidence establishes a Class U trigger.
This annex applies to evaluations using the CertOf default TQ profile. If the requirements in this annex are not met, the evaluation report shall not express the result as CertOf default TQ-A, TQ-B, TQ-C, TQ-D or TQ-U.
| Configuration item | Requirement |
|---|---|
| STU-COUNT-ID | The report shall record the STU counting configuration identifier, version, applicable language, tool or rule name, configuration date and executor. |
| Languages stably segmented by spaces | Count by word unit by default. Consecutive numbers, dates, amounts, identifiers or email addresses are usually recorded as one STU. Punctuation is not counted as a separate STU unless it is itself an evaluated field. |
| Languages not stably segmented by spaces | Use a CertOf-published or project-locked tokenizer, dictionary, rule set or manual segmentation rule, and record the version. If the version or rule is not locked, the CertOf default TQ thresholds shall not be applied. |
| Manual segmentation | Use only when tool segmentation is unavailable or the project expressly requires it. The report shall record segmentation principles, review method and dispute-handling basis. |
| Mixed-language text | Record the STU counting method for the primary language and embedded languages separately. The evaluator shall not alter the default TQ class by selecting a favorable counting method. |
| Item | Example result |
|---|---|
| Object | One-page certificate source document and one-page translated copy. |
| Evidence | ES-FULL. Official source image and official translation PDF are available. |
| TQ | Full text evaluated under an ISO 5060:2024-compatible configuration. Key-field errors: 0. Critical: 0. Major: 0. S <= 2. |
| VE | VE-A. Layout, hierarchy, major objects, reading path and page relationship are highly equivalent. The output is clearly marked as a translated copy. |
| IM | IM-A. Names, dates, certificate number, issuer, signature/seal relationship and page structure are correctly mapped. |
| Class U check | No Class U trigger. |
| Overall result | Class A. |
This example is informative. It does not create additional requirements beyond the normative clauses of this document.
| Area | Questions |
|---|---|
| Evidence | Does the evaluator have official source and official translation output? If not, is the result limited to visible evidence, sample evidence or NR? |
| Format and carrier | Are input format, output format, assessment carrier and editable status recorded? |
| TQ | Was text quality evaluated under an ISO 5060:2024-compatible configuration? Are error categories, severity, penalties or weights, STU counting method, actual STU, Ne, normalized score, rating thresholds and TQ mapping recorded? Does any TQ-U trigger exist? |
| VE | Does the output preserve or restore layout, hierarchy, major visual objects and reading path? If VE-A uses restoration, is the restoration based on verified visual basis? |
| IM | Are labels, values, rows, columns, entities, sequence, references and non-text visual objects correctly mapped? |
| Translated-copy distinction | If the output is a translated copy, is it distinguishable from the original by translation marking or equivalent means? |
| Class U | Does any non-conforming trigger in Annex A apply? |
| Overall result | Does the overall Class X follow the lowest-dimension principle and evidence limitations? |
This annex records pre-publication reference verification only. It does not add normative references. Complete verification notes may be maintained in project review files.
| Reference | Status recorded on 2026-05-28 |
|---|---|
| ISO 5060:2024 | The official ISO page identifies the document as Published, Edition 1, with publication date 2024-02 and reference number ISO 5060:2024. |