On this page本文目录
What Is Semantic Mapping?什么是语义映射？
For the full topic map and the neighboring methods that support this workflow, continue with the complete data integration guide.
如需查看完整主题结构以及支撑本流程的相邻方法，请继续阅读完整的数据集成指南。
Semantic mapping is the documented correspondence between concepts in different vocabularies, taxonomies, ontologies, or data models, with enough context to state whether their meanings are exact, close, broader, narrower, related, conditional, or not validly matched. It explains what two terms mean—not merely where two fields sit or how a value is converted.
语义映射是在不同词表、分类体系、本体或数据模型之间记录概念对应关系，并提供足够上下文，说明两者含义是精确、接近、更宽、更窄、相关、条件成立，还是不存在有效匹配。它解释两个术语究竟表示什么，而不只是字段位于何处或值如何转换。
A source system may say “client,” a master vocabulary may say “customer,” and a regulatory model may distinguish account holder, beneficiary, and purchaser. A name-only match would erase those distinctions. A semantic map records stable identifiers, definitions, scope, relationship type, direction, conditions, provenance, reviewer, and version so a downstream consumer can interpret the correspondence correctly.
源系统可能使用“client”，主词表使用“customer”，而监管模型还会区分账户持有人、受益人和购买者。只按名称匹配会抹去这些区别。语义映射记录稳定标识、定义、范围、关系类型、方向、条件、来源、评审人和版本，使下游使用者能够正确解释对应关系。
The output may be a governed table, graph, SKOS mapping, FHIR ConceptMap, ontology axiom, or a project-specific mapping registry. The representation should match the deployed system. A semantic map is a specification and evidence artifact; it does not by itself transform records, grant access, prove data quality, or guarantee semantic interoperability.
产物可以是受治理的表格、图、SKOS映射、FHIR ConceptMap、本体公理或项目专用映射登记表，表达形式应匹配实际部署系统。语义映射是规范与证据产物；它本身不会转换记录、授予访问、证明数据质量或保证语义互操作。
Choose Semantic Relationship Types Without Overstating Equivalence选择语义关系类型时不要夸大等价性
Use the relationship vocabulary of the standard or application you actually deploy. W3C SKOS provides exact, close, broader, narrower, and related mapping properties for concepts in different schemes. Its formal behavior matters: exact match is transitive, close match is not, and exact match conflicts with hierarchical or related mappings for the same pair. These are SKOS rules, not universal labels that can be copied into every platform unchanged.
应使用实际部署标准或应用的关系词表。W3C SKOS为不同概念体系中的概念提供精确、接近、更宽、更窄和相关等映射属性，其形式行为很重要：精确匹配具有传递性，接近匹配不具有传递性，同一对概念的精确匹配与层级或相关映射相冲突。这些是SKOS规则，并非可以原样复制到所有平台的通用标签。
| Relationship关系 | Use when适用条件 | Primary test主要测试 |
|---|---|---|
| Exact/equivalent精确/等价 | Meanings are interchangeable for the declared scope在声明范围内含义可互换 | Try counterexamples from both directions从两个方向寻找反例 |
| Close/overlapping接近/重叠 | Most intended use aligns, but material exceptions remain多数预期用途一致，但仍有重要例外 | Document the non-interchangeable cases记录不可互换的情况 |
| Broader/narrower更宽/更窄 | One concept's scope contains the other一个概念的范围包含另一个 | Check direction and information loss检查方向与信息损失 |
| Related相关 | Concepts are associated but neither equivalent nor hierarchical概念有关联但既不等价也非层级关系 | Prove that the link supports the use case证明关联能够支持当前用途 |
| Conditional/no match条件/无匹配 | Context determines the target, or no defensible target exists上下文决定目标，或不存在可辩护目标 | Test all conditions and unmapped handling测试全部条件与未映射处理 |
HL7 FHIR's ConceptMap illustrates why direction and context matter: a source concept can have multiple targets, a relationship can state that no valid mapping exists, and reverse validity cannot be assumed. Use such domain standards only where applicable, while keeping the general lesson—context belongs in the mapping.
HL7 FHIR的ConceptMap说明了方向与上下文为何重要：一个源概念可以对应多个目标，关系可以声明不存在有效映射，反向有效性也不能被默认。只在适用领域使用这类行业标准，同时保留一般性原则——上下文必须进入映射。
How to Create a Semantic Mapping Step by Step如何逐步创建语义映射
- Define the decision and scope. Name the consuming query, integration, exchange, search, or analysis; source and target schemes; direction; jurisdiction; and acceptable information loss.定义决定与范围。明确使用映射的查询、集成、交换、搜索或分析，以及源和目标体系、方向、司法辖区与可接受的信息损失。
- Freeze the input versions. Export stable IDs, labels, definitions, hierarchy, synonyms, status, and effective dates. Do not map against an unnamed “latest” vocabulary.冻结输入版本。导出稳定ID、标签、定义、层级、同义词、状态和生效日期，不要针对未命名的“最新版”词表开展映射。
- Profile concepts and usage. Inspect documentation and approved representative values or queries. Labels can suggest candidates, but definitions and actual use reveal scope.剖析概念及其使用。检查文档以及获准的代表性值或查询。标签可以提出候选，但定义和真实使用才能揭示范围。
- Generate candidates. Use exact identifiers, synonyms, hierarchy, lexical similarity, embeddings, external references, or rules as discovery aids. Retain method and score as evidence, not approval.生成候选。使用精确标识、同义词、层级、词法相似度、embedding、外部参考或规则辅助发现；保留方法与分数作为证据，而不是批准。
- Classify relationships. Compare definitions, inclusions, exclusions, granularity, units, time, population, and examples. Record exact, close, broader, narrower, related, conditional, or no-match only under an explicit model.分类关系。比较定义、包含与排除范围、粒度、单位、时间、群体和示例；只在明确模型下记录精确、接近、更宽、更窄、相关、条件或无匹配。
- Review ambiguity and impact. Ask domain owners to examine one-to-many, many-to-one, sensitive, high-frequency, and lossy mappings. Separate proposed, approved, rejected, and retired states.评审歧义与影响。让业务域负责人检查一对多、多对一、敏感、高频及有损映射，并区分提议、批准、拒绝和退役状态。
- Validate, publish, and monitor. Run structural, logical, coverage, example, query, and regression tests; publish the version with provenance and rollback instructions; monitor unmapped and changed concepts.验证、发布并监控。运行结构、逻辑、覆盖、示例、查询和回归测试；连同来源与回滚说明发布版本；监控未映射及变化概念。
Automate repeatable candidate generation and regression checks, but keep review authority explicit. The costliest errors are often plausible matches that pass syntax checks while changing a business definition.
可以自动化可重复的候选生成和回归检查，但评审权限必须明确。代价最高的错误往往是看似合理、通过语法检查，却改变业务定义的匹配。
Semantic Mapping Example: Customer Status Vocabularies语义映射示例：客户状态词表
Hypothetical example: this scenario illustrates the method and is not an InfiniSynapse customer case. A CRM uses ACTIVE, PAUSED, and CLOSED. A billing system uses CURRENT, DELINQUENT, and TERMINATED. An analytics vocabulary defines Engaged Customer, At-Risk Customer, and Former Customer. The labels look similar, but the source concepts describe different dimensions: service lifecycle, payment standing, and analytical behavior.
假设示例：此场景用于说明方法，并非InfiniSynapse客户案例。CRM使用ACTIVE、PAUSED和CLOSED；计费系统使用CURRENT、DELINQUENT和TERMINATED；分析词表定义Engaged Customer、At-Risk Customer和Former Customer。标签看起来相似，但源概念描述的是不同维度：服务生命周期、付款状态与分析行为。
| Candidate候选 | Decision决定 | Reason and test理由与测试 |
|---|---|---|
| ACTIVE → Engaged Customer | Reject as exact; conditional candidate拒绝精确关系；保留条件候选 | An active account may have no recent engagement; test inactivity examples活跃账户可能近期没有互动；测试无互动示例 |
| DELINQUENT → At-Risk Customer | Related, not equivalent相关，但不等价 | Delinquency is one risk signal, not the complete analytical concept欠款只是风险信号之一，并非完整分析概念 |
| CLOSED → Former Customer | Conditional narrower candidate条件成立的更窄候选 | Closure reason and effective date determine whether the person is former关闭原因与生效日期决定其是否属于曾有客户 |
| PAUSED → ? | No match until purpose is known用途明确前无匹配 | Pause may be voluntary, administrative, or temporary failure暂停可能是自愿、行政处理或临时失败 |
The safe implementation keeps the original statuses and maps them only for a declared analysis using additional evidence such as recent activity, closure reason, and payment state. Collapsing all three source vocabularies into one column would be a modeling decision with information loss, not proof of semantic equivalence.
安全实现会保留原始状态，只在声明的分析用途下结合近期活动、关闭原因和付款状态等额外证据进行映射。把三个源词表全部压入一个字段，是会造成信息损失的建模决定，而不是语义等价的证明。
Semantic Mapping vs Data Mapping, Schema Mapping, and Semantic Layers语义映射与数据映射、Schema映射及语义层的区别
| Practice工作 | Primary question主要问题 | Typical output典型产物 |
|---|---|---|
| Semantic mapping语义映射 | How do concepts align in meaning and context?概念在含义与上下文上如何对齐？ | Concept correspondences with typed relations and evidence带关系类型与证据的概念对应 |
| Data mapping数据映射 | Which source element populates which target under what transformation?哪个源元素按什么转换填充哪个目标？ | Source-to-target field and transformation specification源到目标字段及转换规范 |
| Schema mappingSchema映射 | How do structures, paths, types, keys, and constraints correspond?结构、路径、类型、键与约束如何对应？ | Formal structural correspondences or translation rules形式化结构对应或翻译规则 |
| Ontology alignment本体对齐 | Which entities and axioms correspond across ontologies?不同本体中的哪些实体与公理对应？ | Alignment set; may be broader and more formal对齐集合；可能范围更广且更形式化 |
| Entity resolution实体消解 | Do records refer to the same real-world thing?不同记录是否指向同一现实对象？ | Matched entity identities and confidence/evidence已匹配实体身份及置信度/证据 |
| Semantic layer语义层 | How do users query governed business entities, dimensions, and measures?用户如何查询受治理的业务实体、维度与指标？ | Consumption interface and reusable business definitions使用接口与可复用业务定义 |
These artifacts can depend on one another. Semantic decisions may guide a field map; schema mappings may carry the values; a semantic layer may expose the governed result. They should remain traceable rather than being compressed into one ambiguous spreadsheet. Read the separate InfiniSynapse guides to identify data sources and their context and understand the role of a semantic layer.
这些产物可以相互依赖：语义决定可指导字段映射，Schema映射可承载值，语义层可暴露受治理结果。它们应保持可追溯，而不是被压进一张含糊表格。可阅读InfiniSynapse关于识别数据源及其上下文和理解语义层作用的独立指南。
How to Validate Semantic Mappings Before Release发布前如何验证语义映射
Validation must test meaning and operation. A syntactically valid graph can still contain a harmful correspondence, while a correct domain decision can fail because an identifier, direction, or version is wrong.
验证必须同时测试含义与运行。语法有效的图仍可能包含有害对应；正确的业务域决定也可能因为标识、方向或版本错误而失败。
| Layer层次 | Checks检查内容 | Evidence证据 |
|---|---|---|
| Structural结构 | IDs resolve, versions exist, required fields and directions are presentID可解析、版本存在、必填项与方向完整 | Schema/shape validation reportSchema/形状验证报告 |
| Logical逻辑 | No forbidden relation combinations, cycles, contradictions, or invalid cardinality无禁止的关系组合、环路、矛盾或无效基数 | Reasoner or rule-test results推理器或规则测试结果 |
| Semantic语义 | Definitions, inclusions, exclusions, examples, and counterexamples support the relation定义、包含、排除、示例与反例支持关系 | Domain review with rationale带理由的业务域评审 |
| Coverage覆盖 | Expected concepts mapped; unknown, retired, and no-match states counted separately预期概念已映射；未知、退役和无匹配分开计数 | Coverage and exception report覆盖与异常报告 |
| Operational运行 | Representative queries, translations, filters, aggregates, and downstream interpretations remain correct代表性查询、翻译、过滤、聚合与下游解释保持正确 | Golden cases and regression results黄金案例与回归结果 |
Include adversarial cases: similar labels with different definitions, deprecated targets, missing context, one-to-many branches, unexpected languages, and vocabulary changes. Measure proposed, approved, rejected, retired, no-match, and unreviewed states separately; a single “mapping coverage” percentage can conceal risk.
要包含刁钻案例：标签相似但定义不同、目标已弃用、上下文缺失、一对多分支、意外语言以及词表变化。应分别统计提议、批准、拒绝、退役、无匹配与未评审状态；单一“映射覆盖率”可能掩盖风险。
Use InfiniSynapse to Examine Evidence for Semantic Mapping使用InfiniSynapse检查语义映射证据
InfiniSynapse's public product language describes joint analysis across approved sources such as databases, files, and documents, with natural-language analysis. After a team has defined the mapping question and authorized the inputs, that analysis surface can help domain reviewers compare definitions, approved samples, frequency summaries, and edge cases across connected sources.
InfiniSynapse公开产品语言描述了对数据库、文件和文档等获准来源进行联合分析，并支持自然语言分析。在团队定义映射问题并授权输入后，这一分析界面可以帮助业务域评审人比较已连接来源中的定义、获准样例、频率汇总与边界情况。
Prepare source and target vocabulary exports with stable IDs and versions; definitions and scope notes; approved representative examples; candidate relationships; known no-match cases; and specific questions such as “which source statuses contradict this exact-match proposal?” Use read-only access and exclude sensitive values unless explicitly authorized.
请准备带稳定ID和版本的源与目标词表导出、定义与范围说明、获准的代表性示例、候选关系、已知无匹配情况，以及“哪些源状态与这条精确匹配提议相矛盾”等具体问题。使用只读访问；除非明确获准，否则排除敏感值。
Do not describe InfiniSynapse as automatically creating ontologies, determining semantic equivalence, approving mappings, serving a terminology API, managing vocabulary releases, rewriting production queries, or deploying integration logic. Those remain domain, architecture, engineering, security, and governance responsibilities. Preserve approved relationships in the system of record and test the implemented behavior independently.
不要把InfiniSynapse描述为会自动创建本体、确定语义等价、批准映射、提供术语API、管理词表发布、改写生产查询或部署集成逻辑。这些仍是业务域、架构、工程、安全与治理责任。应在事实系统中保存获批关系，并独立测试实施行为。
Bring versioned vocabularies, definitions, candidate relations, representative examples, and review questions. Use InfiniSynapse to explore evidence across approved connected sources; keep approval and deployment in your governed systems.
请准备版本化词表、定义、候选关系、代表性示例与评审问题。使用InfiniSynapse探索获准连接来源中的证据；批准与部署仍应留在受治理系统中。
Analyze approved connected sources分析获准的已连接来源Semantic Mapping FAQ语义映射常见问题
What is semantic mapping?
什么是语义映射？
Semantic mapping is the documented correspondence between concepts in different vocabularies, taxonomies, ontologies, or data models, including the relationship type, direction, context, evidence, and limits needed to preserve intended meaning.
语义映射是在不同词表、分类体系、本体或数据模型之间记录概念对应关系，并包含关系类型、方向、上下文、证据与限制，以保留预期含义。
How is semantic mapping different from data mapping?
语义映射与数据映射有什么区别？
Data mapping usually specifies which source element populates which target and which transformation applies. Semantic mapping asks whether the source and target concepts mean the same thing, overlap, or have broader, narrower, related, conditional, or no valid correspondence.
数据映射通常规定哪个源元素填充哪个目标以及应用什么转换；语义映射则判断源概念与目标概念是含义相同、部分重叠，还是更宽、更窄、相关、条件成立或根本不存在有效对应。
What relationship types should a semantic map record?
语义映射应记录哪些关系类型？
Use the relationship vocabulary defined by the deployed standard or governance model. Common distinctions include exact, close, broader, narrower, related, conditional, and no-match, but these labels are not automatically interchangeable across standards.
应使用实际部署标准或治理模型定义的关系词表。常见区分包括精确、接近、更宽、更窄、相关、条件映射和无匹配，但不同标准中的这些标签不能自动视为可互换。
How do you validate semantic mappings?
如何验证语义映射？
Validate definitions, scope, direction, cardinality, representative and adversarial examples, unmapped concepts, context-dependent cases, logical consistency, downstream query behavior, domain-owner review, provenance, and regression results after either vocabulary changes.
应验证定义、范围、方向、基数、代表性与刁钻示例、未映射概念、依赖上下文的情况、逻辑一致性、下游查询行为、业务域负责人评审、来源记录，以及任一词表变化后的回归结果。
Authoritative Sources and Next Steps权威来源与下一步
Use the specification that governs the system you deploy. W3C SKOS defines mapping properties and their formal relationships; the W3C semantic interoperability note explains why shared vocabularies or explicit correspondences are needed; and HL7 FHIR ConceptMap demonstrates directional, context-dependent concept mappings, including multiple targets and no valid match. The FHIR source is an industry example, not a universal requirement for non-healthcare projects.
应使用治理实际部署系统的规范。W3C SKOS定义映射属性及其形式关系；W3C语义互操作说明解释为何需要共享词表或明确对应；HL7 FHIR ConceptMap展示有方向、依赖上下文的概念映射，包括多目标和无有效匹配。FHIR来源是行业示例，并非非医疗项目的通用要求。
- W3C SKOS Reference: mapping propertiesW3C SKOS参考：映射属性
- W3C note on semantic integration and interoperabilityW3C语义集成与互操作说明
- HL7 FHIR ConceptMap scope and usageHL7 FHIR ConceptMap范围与用法
- InfiniSynapse guide to data sourcesInfiniSynapse数据源指南
- InfiniSynapse guide to semantic layersInfiniSynapse语义层指南
- InfiniSynapse guide to data lineageInfiniSynapse数据血缘指南