知识图谱导论：起源、定义与演进 Introduction to Knowledge Graphs: Origins, Definitions, and Evolution

从20世纪90年代万维网的诞生到21世纪初语义Web的提出，再到2012年Google Knowledge Graph的横空出世，知识图谱（Knowledge Graph）作为信息管理与人工智能交叉领域的核心概念，经历了三十余年的演进历程。在当前以大语言模型（Large Language Models, LLMs）和生成式人工智能（Generative AI）为代表的新一轮AI革命中，知识图谱凭借其结构化的知识表示、明确的语义关联和可解释的推理能力，正在重新定位其在现代智能系统中的战略价值。理解知识图谱的起源、定义与技术演进，不仅是掌握这一领域基础知识的必要前提，更是把握人工智能未来发展脉络的关键所在。

The evolution from the World Wide Web's emergence in the 1990s to the Semantic Web's proposal in the early 2000s, and subsequently the launch of Google Knowledge Graph in 2012, represents a journey spanning over three decades for the concept of Knowledge Graph as a cornerstone at the intersection of information management and artificial intelligence. Within the current AI revolution characterized by Large Language Models (LLMs) and Generative AI, knowledge graphs—leveraging their structured knowledge representation, explicit semantic connections, and interpretable reasoning capabilities—are strategically repositioning themselves within modern intelligent systems. Comprehending the origins, definitions, and technological evolution of knowledge graphs constitutes not merely a prerequisite for mastering this domain's fundamentals, but also a critical key to understanding the trajectory of artificial intelligence's future development.

---

1. 从万维网到语义Web：知识图谱的思想渊源

1. From the World Wide Web to the Semantic Web: Intellectual Origins of Knowledge Graphs

1.1 万维网的诞生与信息爆炸

1.1 The Birth of the World Wide Web and the Information Explosion

1989年，欧洲核子研究组织（CERN）的Tim Berners-Lee提出了万维网（World Wide Web, WWW）的构想，这一革命性的技术彻底改变了人类获取和共享信息的方式。Berners-Lee设计的万维网架构基于三个核心标准：统一资源定位符（Uniform Resource Locator, URL）作为信息资源的地址标识、超文本传输协议（HyperText Transfer Protocol, HTTP）作为客户端与服务器之间的通信协议、以及超文本标记语言（HyperText Markup Language, HTML）作为描述网页内容的标记语言。这三项技术的有机结合，使得全球范围内的文档可以通过超链接（Hyperlink）相互连接，形成了一个人类历史上前所未有的巨型互联文档网络。

In 1989, Tim Berners-Lee, working at CERN (the European Organization for Nuclear Research), proposed the concept of the World Wide Web—a revolutionary technology that fundamentally transformed how humanity accesses and shares information. Berners-Lee's web architecture was built upon three core standards: Uniform Resource Locators (URLs) for addressing information resources, the HyperText Transfer Protocol (HTTP) for client-server communication, and HyperText Markup Language (HTML) for describing web page content. The organic integration of these three technologies enabled documents across the globe to interconnect through hyperlinks, forming an unprecedented giant network of interconnected documents in human history.

万维网的指数级增长令人惊叹。根据Internet World Stats的统计数据，全球互联网用户从2000年的3.61亿增长到2023年的超过53亿，普及率从不足6%攀升至约67%。与此同时，网页数量也从最初的数千个增长到估计超过13万亿个。然而，这种爆炸式增长也带来了严峻的信息检索挑战：当用户面对海量网页时，如何快速、准确地找到真正所需的信息？传统的基于关键词匹配的搜索引擎虽然能够在一定程度上缓解这一问题，但其本质局限——无法理解查询意图与文档内容的语义——始终制约着信息检索效果的进一步提升。

The exponential growth of the World Wide Web has been staggering. According to Internet World Stats, global internet users grew from 361 million in 2000 to over 5.3 billion in 2023, with internet penetration rising from under 6% to approximately 67%. Simultaneously, the number of web pages has grown from thousands in the early days to an estimated over 13 trillion. However, this explosive growth has also introduced severe information retrieval challenges: when users face massive amounts of web pages, how can they quickly and accurately locate the information they truly need? Traditional keyword-based search engines, while alleviating this problem to some extent, suffer from an inherent limitation—their inability to understand the semantic meaning of both query intent and document content—which has perpetually constrained further improvements in information retrieval effectiveness.

1.2 语义Web的愿景与技术框架

1.2 The Vision of the Semantic Web and Its Technical Framework

面对万维网的信息过载困境，Tim Berners-Lee于1998年正式提出了语义Web（Semantic Web）的概念，旨在为Web内容赋予机器可理解的语义信息，从而实现智能化的信息检索与知识处理。Berners-Lee将语义Web描绘为当前万维网的一个扩展，在这个扩展中，信息被赋予了明确的含义，使得计算机与人类能够更加高效地协同工作。2001年，Berners-Lee与国际万维网联盟（W3C）的同事们在《科学美国人》杂志上发表了一篇具有里程碑意义的文章《语义Web：一个新兴的计算机与信息科学新领域》，系统阐述了语义Web的愿景、架构与技术路线。

Confronted with the information overload dilemma of the World Wide Web, Tim Berners-Lee formally introduced the concept of the Semantic Web in 1998, aiming to endow web content with machine-interpretable semantic information, thereby enabling intelligent information retrieval and knowledge processing. Berners-Lee depicted the Semantic Web as an extension of the current web where information is given well-defined meaning, enabling computers and humans to work together more effectively. In 2001, Berners-Lee and colleagues from the World Wide Web Consortium (W3C) published a landmark article titled "The Semantic Web: A New Form of Web Content That is Meaningful to Computers Will Unfold a New Form of Current Knowledge" in Scientific American, systematically articulating the vision, architecture, and technical roadmap of the Semantic Web.

语义Web的技术架构遵循经典的分层模型（Layered Model），这一模型借鉴了计算机网络中的OSI参考模型的设计思想，由底层到顶层依次包括：Unicode与URI层负责字符编码与资源标识；XML与命名空间层提供语法基础；RDF层描述资源及其属性；本体层（OWL）表达概念与关系；逻辑层与证明层支持推理与验证；信任层则确保应用的安全性。这一分层架构的核心理念在于：每一层都建立在下一层之上，并通过引入新的语义表达能力来增强整个系统的知识表示与推理水平。

The Semantic Web's technical architecture follows the classic layered model, drawing design inspiration from the OSI reference model in computer networking. From bottom to top, this model includes: Unicode and URI layers for character encoding and resource identification; XML and namespaces for syntactic foundations; RDF for describing resources and their properties; the ontology layer (OWL) for expressing concepts and relationships; logic and proof layers for reasoning and verification; and the trust layer ensuring application security. The core philosophy of this layered architecture lies in: each layer builds upon the one below it, progressively enhancing the system's knowledge representation and reasoning capabilities through introducing new semantic expressiveness.

语义Web运动催生了一系列重要的技术标准与基础设施，其中最具影响力的包括RDF、OWL和SPARQL。资源描述框架（Resource Description Framework, RDF）作为语义Web的数据模型，提供了描述Web资源及其关系的基础语法；Web本体语言（Web Ontology Language, OWL）作为语义Web的本体描述语言，提供了丰富的语义表达能力；SPARQL协议与RDF查询语言（SPARQL Protocol and RDF Query Language）则为RDF数据提供了标准化的查询接口。这些技术标准的制定，为后续知识图谱的发展奠定了坚实的理论基础与工程实践基础。

The Semantic Web movement gave birth to several pivotal technical standards and infrastructures, with the most influential being RDF, OWL, and SPARQL. The Resource Description Framework (RDF), serving as the Semantic Web's data model, provides a foundational syntax for describing web resources and their relationships; the Web Ontology Language (OWL), as the Semantic Web's ontology description language, offers rich semantic expressiveness; and SPARQL (SPARQL Protocol and RDF Query Language) provides a standardized query interface for RDF data. The establishment of these technical standards laid a solid theoretical foundation and engineering practice base for the subsequent development of knowledge graphs.

1.3 语义网络：知识表示的早期探索

1.3 Semantic Networks: Early Explorations in Knowledge Representation

在语义Web概念正式提出之前，人工智能领域已经对语义网络（Semantic Network）进行了数十年的研究与探索。语义网络作为一种知识表示方法，其基本思想是用图（Graph）的结构来表达知识：节点（Node）表示概念或实体，弧（Arc）或边（Edge）表示概念之间的关系。这种表示方法最早可以追溯到Quillian在1966年提出的语义记忆模型（Teachable Language Comprehender, TLC），该模型尝试用互相连接的节点网络来模拟人类语言的语义组织方式。

Prior to the formal proposal of the Semantic Web concept, the field of artificial intelligence had already conducted decades of research and exploration into semantic networks. Semantic networks, as a knowledge representation method, fundamentally employ graph structures to express knowledge: nodes represent concepts or entities, while arcs or edges represent relationships between concepts. This representation approach traces its roots back to Quillian's 1966 proposal of the semantic memory model (Teachable Language Comprehender, TLC), which attempted to simulate the semantic organization of human language through interconnected node networks.

在后续的发展中，语义网络被广泛应用于自然语言处理（Natural Language Process