⌕ ESC
↑↓ 选择↵ 打开esc 关闭⌘K 唤起
← 返回速读报告 Drizzle Digest · NO.213 · 全文

What is ontology (preview)? — Microsoft Fabric IQ

频道: Microsoft Learn
原文: https://learn.microsoft.com/en-us/fabric/iq/ontology/overview
原文语言: en
统计: 共 15 轮

[¶1]

What is ontology (preview)?

The ontology (preview) item (part of the Fabric IQ (preview) workload) digitally represents the enterprise vocabulary and semantic layer that unifies meaning across domains and OneLake sources. It defines enterprise concepts as entity types (like Customer), properties (like a Customer's name and email), and relationships (like Customer places Order), while clarifying the constraints of these terms. After defining your ontology, bind the entity type definitions to real data, so downstream tools can share the same language. Both humans and AI agents can use this language for cross-domain reasoning and decision-ready actions.

什么是本体（Ontology，预览版）？

本体（预览版）项（属于 Fabric IQ（预览版）工作负载）以数字化方式呈现企业词汇与语义层，在各业务域和各 OneLake 数据源之间统一含义。它将企业概念定义为实体类型（entity type，如客户）、属性（property，如客户的姓名和电子邮件）和关系（relationship，如客户下订单），同时明确这些术语的约束条件。定义好本体后，将实体类型定义绑定到真实数据，让下游工具共用同一套语言。无论是人还是 AI 智能体，都可以借助这套语言开展跨域推理，并采取可直接支撑决策的行动。

[¶2]

Ontology in Fabric provides a scaled, secure, and governed shared business model used across teams, agents, and workflows in Fabric IQ. It provides a shared context layer that can be consumed by Fabric agents and Real-Time Intelligence components for consistent reasoning and actions. Ontology works well in situations where you need cross-domain consistency, governance, or AI agent grounding, and you want to reason across processes.

[!INCLUDE Fabric feature-preview-note]

Fabric 中的本体提供一个可规模化、安全且受治理的共享业务模型，供 Fabric IQ 中的各个团队、智能体和工作流共同使用。它提供一个共享的上下文层，Fabric 智能体和 Real-Time Intelligence 组件都可以使用这一层，从而做出一致的推理与行动。当你需要跨域一致性、治理能力或为 AI 智能体提供事实依据（grounding），并且希望跨流程进行推理时，本体尤为适用。

[!INCLUDE Fabric feature-preview-note]

[¶3]

Ontology overview

An ontology is a shared, machine-understandable vocabulary of your business. It's made up of the things in your environment (represented as entity types), their facts (represented as properties on entity types), and the ways they connect (represented as relationships), while offering constraints and rules that keep representations consistent.

You can also think of an ontology like a business context layer, containing:

本体概述

本体是描述你业务的一套共享、机器可理解的词汇。它由你所处环境中的事物（以实体类型表示）、这些事物的事实（以实体类型上的属性表示）以及事物之间的连接方式（以关系表示）构成，同时提供约束和规则，确保这些表示保持一致。

你也可以把本体看作一个业务上下文层，其中包含：

[¶4]

Core concepts: Defining an ontology

An ontology consists of entity types, entity instances, properties, and relationships. This section describes each of these core concepts.

核心概念：定义本体

本体由实体类型、实体实例（entity instance）、属性和关系组成。本节将逐一介绍这些核心概念。

[¶5]

Entity type

An entity type is the reusable logical model of a real world concept (like Shipment, Product, or Sensor). It standardizes the name, description, identifiers, properties, and constraints for that item, so that every team in your business means the same thing when they use a term like "shipment." By elevating the concept above any single table, entity types eliminate conflicting column level definitions across sources. They provide a single point to attach properties, relationships, and labels that downstream tools can use to improve semantics across tables and models.

实体类型

实体类型是现实世界中某个概念（如运单、产品或传感器）的可复用逻辑模型。它对该概念的名称、描述、标识符、属性和约束进行标准化，让企业里的每个团队在使用“运单”这样的术语时，指的都是同一个东西。实体类型把概念提升到任何单张表之上，从而消除各数据源之间相互冲突的列级定义。它们提供了一个统一的挂载点，用来附加属性、关系和标签，下游工具可以借助这些信息改进跨表、跨模型的语义。

[¶6]

Entity instance

An entity instance is a concrete occurrence of an entity type, populated from data bindings (like a semantic row). Entity instances keep track of which source created them and when they were true, and can participate in relationships. Entity instances turn your raw data into standardized business objects that all your tools and AI agents can understand in the same way.

实体实例

实体实例是实体类型的一个具体个体，由数据绑定填充而来（相当于一行带语义的数据）。实体实例会记录自己由哪个数据源创建、在何时有效，并且可以参与关系。实体实例把原始数据转化为标准化的业务对象，你的所有工具和 AI 智能体都能以相同的方式理解这些对象。

[¶7]

Property

A property is a named fact about an entity, with a declared data type. It can contain bindings to source data and semantic annotations (like an identifier or metadata attributes). Properties improve semantics by enforcing consistent types, units, and naming, and by enabling rules and quality checks at the concept level.

属性

属性是关于实体的一项具名事实，并带有声明的数据类型。它可以包含与源数据的绑定，以及语义注释（如标识符或元数据特性）。属性通过强制统一类型、单位和命名，并支持在概念层面执行规则和质量检查，来增强语义。

[¶8]

Relationship

A relationship is a typed, directional link between entity types or instances. Relationships can have attributes (like distance, confidence, or effectiveAt) and cardinality rules defining how many items can be related (for example, one Customer having many Orders). Relationships make context explicit and reusable for how things connect, enabling traversal, dependency analysis, rule based inference, and clearer answers to business questions without custom join logic.

关系

关系是实体类型之间或实例之间有类型、有方向的链接。关系可以带有特性（如距离、置信度或 effectiveAt），以及用于规定可关联数量的基数规则（例如，一个客户可以有多个订单）。关系让“事物之间如何连接”这层上下文变得显式且可复用，从而支持遍历、依赖分析和基于规则的推理，无需编写自定义联接逻辑，就能更清晰地回答业务问题。

[¶9]

Core concepts: Your data in the ontology

After you define an ontology, you can bind it to your data to visualize and query the data in the context of your ontology. Read about each core data concept in the following sections.

核心概念：本体中的数据

定义好本体后，你可以将其绑定到数据，从而在本体的上下文中对数据进行可视化和查询。以下各节将分别介绍各个核心数据概念。

[¶10]

Data binding

Data binding connects your ontology's definitions (including entity types, properties, and relationships) to concrete data living in OneLake, including lakehouse tables, eventhouse streams, and Power BI semantic models. A data binding describes data types, identity keys, how columns map to properties, and how keys map to relationships across multiple data sources. By enabling schema evolution rules, data quality checks (based on things like nullability, ranges, and uniqueness), and provenance at the concept layer, bindings turn raw rows and events into governed business objects. Data binding ensures your data has consistent meaning, follows the same rules, and tracks where it came from across different data sources.

数据绑定

数据绑定将本体中的定义（包括实体类型、属性和关系）连接到存放在 OneLake 中的具体数据，包括 lakehouse 表、eventhouse 流和 Power BI 语义模型。数据绑定描述了数据类型、标识键、列如何映射到属性，以及键如何在多个数据源之间映射为关系。通过在概念层启用 schema 演进规则、数据质量检查（基于可空性、取值范围、唯一性等）和来源追溯，绑定把原始的数据行和事件转化为受治理的业务对象。数据绑定确保你的数据在不同数据源之间含义一致、遵循相同的规则，并能追踪其来源。

[¶11]

[!INCLUDE refresh-graph-model]

[!INCLUDE refresh-graph-model]

[¶12]

Ontology graph

The ontology graph is a queryable instance graph built from your data bindings and relationship definitions, provided within ontology by Graph in Microsoft Fabric. You can see the graph in the entity type details.

In the graph, nodes are entity instances, and edges are links (either asserted or derived) with metadata attributes. Each node or edge keeps data source lineage and follows a scheduled data refresh. Graphs enable visual exploration of business context, execution of graph algorithms (like paths, centrality, and communities), and rule‑driven inferences. Graphs improve semantics by making relationships first‑class, so context is explicit, queryable, and governed (not buried in join logic).

本体图

本体图（ontology graph）是基于数据绑定和关系定义构建的可查询实例图，由 Graph in Microsoft Fabric 在本体中提供。你可以在实体类型详细信息中查看该图。

在图中，节点是实体实例，边是带有元数据特性的链接（可以是显式声明的，也可以是推导得出的）。每个节点或边都会保留数据源血缘，并按计划刷新数据。借助图，你可以对业务上下文进行可视化探索、运行图算法（如路径、中心性和社区发现），以及执行规则驱动的推理。图把关系提升为一等公民，使上下文变得显式、可查询且受治理（而不是隐藏在联接逻辑中），从而增强语义。

[¶13]

[!INCLUDE refresh-graph-model]

[!INCLUDE refresh-graph-model]

[¶14]

Querying your ontology

Ontology querying lets you ask business-level questions over bound data sources through ontology terminology. Queries start with entity types and allow filtering by properties, traversing relationships, aggregating by time, and other constraints. The ontology layer automatically sends your queries to the most efficient system to get results quickly (such as GQL for Graph in Microsoft Fabric and KQL for Eventhouse). It also includes a Natural Language to Ontology (NL2Ontology) query layer, which converts your natural language questions into structured queries and returns relevant results. This enables you to ask questions using business terms, instead of needing to know the details of how your data is stored in different systems. NL2Ontology queries ensure that filters, joins, units, and validity windows align with the definitions that are published in your ontology.

查询本体

本体查询让你可以借助本体术语，针对已绑定的数据源提出业务层面的问题。查询以实体类型为起点，支持按属性筛选、遍历关系、按时间聚合以及施加其他约束。本体层会自动把查询发送到效率最高的系统，以快速获得结果（例如 Graph in Microsoft Fabric 使用 GQL，Eventhouse 使用 KQL）。它还包含一个自然语言到本体（NL2Ontology）查询层，能把你用自然语言提出的问题转换为结构化查询，并返回相关结果。这样，你就能直接用业务术语提问，而无需了解数据在不同系统中如何存储的细节。NL2Ontology 查询可确保筛选条件、联接、单位和有效时间窗口与本体中已发布的定义保持一致。

[¶15]

Next steps

后续步骤