RDFa
RDFa (Resource Description Framework in Attributes) is a W3C specification that defines a set of attributes for embedding structured, machine-readable data within markup languages such as HTML5, XHTML, and any XML-based format, allowing the expression of RDF (Resource Description Framework) triples directly in the content without altering its human-readable presentation.[1] Developed by the World Wide Web Consortium (W3C), RDFa enables web authors to annotate existing hypertext with semantic metadata, facilitating applications like improved search engine indexing, social media integration, and data interoperability across the web.[2]
The specification originated in the mid-2000s as an extension of XHTML to address the limitations of plain HTML in conveying machine-understandable semantics, with RDFa 1.0 becoming a W3C Recommendation in October 2008 specifically for XHTML. Subsequent evolution led to RDFa 1.1 in 2012, which expanded support to HTML5 and introduced RDFa Core as a language-agnostic foundation, alongside lighter variants like RDFa Lite for simpler use cases.[1] Key attributes include @property for predicates, @resource for objects, @vocab for default vocabularies, and @prefix for namespace mappings, enabling the integration of diverse ontologies such as Dublin Core for metadata or FOAF for social relationships.[2]
RDFa's design emphasizes reuse of existing markup to avoid redundancy, distinguishing it from alternatives like Microdata by its basis in RDF standards and flexibility with multiple vocabularies, which promotes broader semantic web adoption.[1] As of 2015, the latest editions of RDFa Core, HTML+RDFa, and XHTML+RDFa provide processing rules and conformance guidelines, ensuring consistent extraction of structured data by processors.[3] This framework has been instrumental in initiatives like schema.org, where publishers mark up content for enhanced discoverability by search engines and other services.[2]
History
Origins and Early Development
RDFa originated from efforts to integrate the Resource Description Framework (RDF) data model directly into markup languages, allowing structured data to be embedded within human-readable content. In February 2004, Mark Birbeck published the W3C Note "XHTML and RDF", outlining techniques for embedding RDF in XHTML. This was followed in October 2004 by the W3C Note "RDF/A Syntax", which proposed a syntax for layering RDF information onto any XML document using attributes, specifically targeting XHTML to overcome the challenges of traditional RDF/XML serialization in web contexts.[4][5] This initial proposal addressed key limitations, such as the difficulty of validating RDF/XML with XML schemas or DTDs and the awkwardness of using it for inline document metadata without relying on separate external files.[5]
The development of RDFa involved close collaboration between key contributors, including Mark Birbeck and Ben Adida, under the auspices of the W3C Semantic Web Deployment Working Group (SWD-WG), which worked in cooperation with the HTML Working Group.[6] Birbeck and Adida co-authored early documentation, such as the RDFa Primer, emphasizing practical embedding techniques for RDF triples—statements consisting of subject, predicate, and object—directly into HTML/XHTML elements to avoid the need for external RDF files and enable seamless integration of structured data.[7] The primary motivation was to facilitate the expression of RDF within existing web documents, reusing visible content as structured data without duplication, thus promoting broader adoption of semantic technologies on the web.[8]
RDFa's evolution was closely tied to the development of XHTML 2.0, where it was proposed as a dedicated module to extend the language's capabilities for semantic markup. Early drafts of the XHTML 2.0 specification incorporated RDFa concepts, positioning it as a native way to embed RDF in modularized XHTML documents and aligning with the W3C's vision for more expressive hypertext.[9] This integration into XHTML 2.0 drafts marked a pivotal step in transitioning RDF from standalone formats to inline web authoring tools.[10]
Standardization Milestones
The standardization of RDFa began with the publication of the "RDFa in XHTML: Syntax and Processing" as a First Public Working Draft on October 18, 2007.[11] This was followed by its advancement to Last Call Working Draft on February 21, 2008,[12] during which feedback was solicited, and then to Candidate Recommendation status on June 20, 2008, during which implementations were tested to ensure compliance and interoperability.[13]
RDFa 1.0 achieved W3C Recommendation status on October 14, 2008, with the release of both the "RDFa in XHTML: Syntax and Processing" (specifying XHTML+RDFa 1.0) and the accompanying RDFa 1.0 Primer as a W3C Note, establishing the initial core syntax and processing rules for structured data in XHTML.[14] These documents provided the foundational attributes and mechanisms for RDFa, enabling publishers to embed machine-readable data without altering the visual presentation of web pages.
The evolution continued with RDFa 1.1, where the "RDFa Core 1.1" specification was published as a W3C Recommendation on June 7, 2012, extending the syntax beyond XML-based languages like XHTML to support non-XML host languages such as HTML4 and HTML5.[15] This version introduced greater flexibility, including RDFa Lite 1.1 as a simplified subset for easier adoption. Accompanying syntax documents, such as "XHTML+RDFa 1.1," were also recommended on the same date.[16]
The RDFa 1.1 Primer, providing guidance on implementation and use cases, received its last major update as the Third Edition on March 17, 2015, reflecting refinements based on community feedback while maintaining compatibility with prior versions.[2]
Post-2012, the W3C RDFa Working Group addressed minor issues through errata documents, including corrections to RDFa Core 1.1, RDFa Lite 1.1, and the Primer, with ongoing updates tracked as late as 2015 but no substantive revisions or new major versions issued by November 2025.[17][18] This stability has supported widespread implementation without disrupting existing deployments.
Core Concepts
RDF Foundations
RDF (Resource Description Framework) is a W3C standard that provides a framework for representing information on the Web in the form of subject-predicate-object triples, enabling the description of resources and their relationships in a machine-readable way.[19] Each triple asserts a statement about a resource, where the subject identifies the resource being described, the predicate specifies the property or relationship, and the object provides the value or related resource.[20] This triple-based structure forms the foundational data model for RDF, allowing for flexible, interconnected data representation without a fixed schema.[21]
In RDF, subjects, predicates, and objects are primarily identified using Internationalized Resource Identifiers (IRIs), which are a generalization of URIs that support international characters and serve as globally unique names for resources.[19] Subjects and predicates must be IRIs (or, in the case of subjects, blank nodes for anonymous resources), while objects can be either IRIs (denoting resources) or literals (denoting values such as strings, numbers, or dates).[20] Literals differ from resources in that they represent concrete values rather than entities that can be further described, often including optional datatype IRIs or language tags to specify their type or context.[19]
An RDF graph is defined as a set of these triples, forming a directed, labeled graph where subjects and objects act as nodes and predicates as edges.[20] To manage the verbosity of full IRIs in practical use, RDF serializations commonly employ namespaces—base URIs that define vocabularies—and prefixes as abbreviations for those namespaces, facilitating concise expression without altering the underlying abstract model.[19]
RDF 1.1, published as a W3C Recommendation in 2014, formalized the core concepts including IRIs, literals, and graphs, while RDF 1.2, advanced to Recommendation status in 2025, introduced enhancements such as support for triple terms but maintained backward compatibility with the 1.1 model.[19][20] RDFa, as an embedding syntax, aligns with the RDF 1.1 data model, using its attributes to generate conforming triples within host languages.[1]
RDFa Syntax and Attributes
RDFa embeds RDF triples into markup languages using a set of dedicated attributes that specify subjects, predicates, objects, and types within the document's elements. The core attributes include about, which establishes the subject of a statement by referencing a URI or generating a new one from the current document; rel and rev, which define forward and reverse relationships between resources using predicates; property, which links a subject to a literal value or another resource as an object; resource, which specifies a non-triplified object URI; typeof, which assigns an RDF class to the subject via the rdf:type predicate; prefix, which declares mappings from namespace prefixes to full IRIs; and vocab, which sets a default vocabulary IRI for interpreting terms without explicit prefixes.[1]
These attributes operate on host language elements, such as XHTML or HTML, allowing RDF data to be interwoven with visible content. For instance, in the markup <div about="http://example.org/person" typeof="foaf:Person" property="foaf:name">[Alice](/page/Alice)</div>, the about attribute sets the subject, typeof asserts the class, and property creates a triple with the element's text content as the literal object. Similarly, <a rel="foaf:knows" href="http://example.org/bob">knows Bob</a> uses rel to generate a triple where the current subject relates to the href target via the foaf:knows predicate. The rev attribute inverts this, making the linked resource the subject instead.[1]
To compactly reference Internationalized Resource Identifiers (IRIs), RDFa employs Compact URIs (CURIEs), which take the form prefix:local (e.g., foaf:name) or the bracketed [prefix:local] for safe contexts like attribute values. CURIEs resolve to full IRIs by substituting the prefix with its mapped IRI, falling back to the vocab IRI if no prefix is specified, or using absolute IRIs directly. This mechanism, along with the prefix attribute for on-the-fly declarations (e.g., prefix="foaf: http://xmlns.com/foaf/0.1/"), enables concise expression of RDF predicates and types without repetitive full URIs.[1]
Default profiles provide initial context for interpretation, predefined by the host language and including common prefix mappings and terms, such as those in the RDFa 1.1 profile at https://www.w3.org/2011/rdfa-context/rdfa-1.1, which maps terms like license to Creative Commons predicates. Inheritance rules propagate context from parent to child elements: the subject defaults to the parent's if not overridden by about; predicates and terms inherit vocabulary mappings; and language or datatype settings cascade unless locally specified. This hierarchical model ensures that nested elements can extend or refine the RDF graph without redundant declarations.[1]
The RDFa processing model generates an RDF graph through sequential evaluation of each element's context, starting from the document's base IRI and an initial empty subject. For each element, the processor first determines the local context by applying attributes like about to set or inherit the subject, typeof to add type triples, and vocab or prefix to update mappings. It then identifies potential triples: for property, a triple forms with the current subject, the property as predicate, and the object derived from the element's content, resource, or child resources (optionally modified by content or datatype attributes for plain literals or typed values); for rel or rev, triples use the subject/object pair with the relation as predicate, completing incomplete triples via nested about or resource. Context evaluation handles scoping, such as bNode generation for blank nodes when no explicit subject is provided, and propagates changes to descendants. The resulting triples form a merged RDF dataset, serialized in formats like Turtle or N-Triples for output, ensuring the embedded data is extractable as a standard RDF graph.[1]
Versions and Profiles
RDFa 1.0
RDFa 1.0 was introduced as a W3C Recommendation on 14 October 2008, comprising the RDFa in XHTML: Syntax and Processing specification along with the XHTML RDFa Attribute Module for integration.[14] This version enabled the embedding of Resource Description Framework (RDF) metadata directly into XHTML documents using a set of attribute extensions, allowing structured data to be expressed alongside human-readable content without duplication.[14] Designed primarily for XHTML 1.1, RDFa 1.0 built upon existing XHTML attributes while introducing new ones to facilitate RDF triples, focusing on XML-based host languages to ensure precise parsing and validation.[14]
The syntax of RDFa 1.0 relies on a full suite of attributes to define subjects, predicates, objects, and datatypes within XHTML elements. Core attributes include rel and rev for specifying relationships (using Compact URI Expressions or CURIEs for brevity), datatype to indicate the type of literal values (defaulting to XML Schema string), about for identifying the subject URI, property for predicates, resource for object URIs, typeof for class declarations, content for overriding visible text, and XHTML-native href and src for links and resources.[14] These attributes must be used in well-formed XML documents conforming to XHTML 1.1 schemas, with processing rules that generate RDF graphs through a step-by-step evaluation of the DOM tree, starting from the root element and inheriting context via xmlns prefixes for namespaces.[14] A key limitation is the strict requirement for XML well-formedness, which precludes compatibility with non-XML HTML parsing algorithms, such as those later defined for HTML5.[14]
Supporting documentation for RDFa 1.0 includes the RDFa Primer 1.0, published as a W3C Note on 14 October 2008, which provides introductory examples of embedding RDFa in XHTML for scenarios like marking up events, people, and publications.[22] The primer emphasizes practical usage, such as using rel="foaf:depiction" on an image element to link it to a person's profile, and explains inheritance rules for generating complete triples.[22] Additionally, a test suite comprising 98 approved cases was developed to validate implementations, covering features like property chaining, XML literals, and CURIE resolution; the 2008 implementation report confirmed that four processors (RDFa2RDFXML, RDFa Distiller, librdfa, and SPREAD) passed all tests, demonstrating robust interoperability for XHTML+RDFa.[23]
RDFa 1.1
RDFa 1.1, published as a W3C Recommendation on 7 June 2012, represents a significant evolution from RDFa 1.0 by decoupling the core syntax and processing rules from specific host languages, enabling broader applicability across various markup formats including non-XML languages.[1] This separation allows host languages to define their own profiles, such as default vocabularies, prefix mappings, and initial contexts, promoting language-independent implementations that enhance internationalization and flexibility in embedding RDF data.[1] Unlike RDFa 1.0, which was tightly coupled to XHTML, the 1.1 core specification focuses on generic XML processing rules, with host-specific adaptations handled in companion documents.[1]
Key enhancements include the introduction of new attributes to streamline RDF expression and improve prefix management. The @inlist attribute, for instance, allows elements to contribute to RDF lists by appending objects to a predicate's collection, with its value ignored and a new list created if none exists, facilitating structured data representations like ordered relations.[1] Prefix handling was refined with the @prefix attribute, which declares compact URIs (CURIEs) in a case-insensitive manner, scoped locally with inner mappings taking precedence, and deprecating the @xmlns: syntax for cleaner integration.[1] Additionally, the @vocab attribute establishes a default vocabulary IRI for undefined terms, supporting language-independent profiles that reduce verbosity in markup.[1]
RDFa 1.1 aligns closely with HTML5 parsing rules, extending support to HTML5 documents in both non-XML (text/html) and XML (application/xhtml+xml) modes, as well as HTML4 and XHTML, while incorporating RDFa attributes like @typeof, @[property](/page/Property), @[resource](/page/Resource), @[content](/page/Content), @about, @rel, @[rev](/page/Rev), @datatype, and @inlist.[3] This compatibility ensures processors can extract RDF triples from HTML5 content using standardized parsing, including features like property copying from parent elements and typed literals via @datetime.[3] SVG integration is also enabled, preserving namespace declarations in XMLLiterals for vector graphics markup, thus broadening RDFa 1.1's utility in multimedia and diagrammatic contexts.[3]
RDFa Lite
RDFa Lite is a minimal subset of the RDFa 1.1 specification, designed to simplify the embedding of structured data in web documents by restricting the available attributes to a core set of five: vocab, typeof, [property](/page/Property), resource, and [prefix](/page/Prefix).[24] This approach eliminates more advanced attributes such as rel, rev, and about, which are part of the full RDFa 1.1 but omitted here to reduce complexity and learning curve for users.[24]
Specified as part of RDFa 1.1 and published as a W3C Recommendation on June 7, 2012, RDFa Lite aims to provide a gentler introduction to structured data markup for web authors who may be unfamiliar with RDF concepts.[25] It targets publishers seeking to express simple machine-readable data, such as basic information about people, events, or products, without needing to master the full RDFa syntax.[24]
The profile is fully upwards compatible with RDFa 1.1, allowing markup created with RDFa Lite to be extended later by incorporating additional attributes from the complete specification.[24] RDFa Lite is particularly well-suited for use with popular vocabularies like schema.org, enabling straightforward integration of common web data types while maintaining compatibility with RDF-based tools.[24] This "lite" design rationale emphasizes widespread adoption by prioritizing ease of use over comprehensive expressiveness, addressing feedback from early RDFa implementations that highlighted barriers for non-experts.[24]
Host Language Integration
XHTML and XML Support
RDFa was integrated into XHTML as a modular extension, with "RDFa in XHTML: Syntax and Processing" becoming a W3C Recommendation on 14 October 2008, defining how RDFa attributes could be used within XHTML Family Markup Languages to embed structured data.[14] This specification positioned RDFa as a superset of XHTML 1.1, allowing authors to annotate documents with RDF triples while maintaining compatibility with the XHTML modularization framework.[14] The integration was developed by the XHTML 2 Working Group and the Semantic Web Deployment Working Group, enabling RDFa to leverage XHTML's XML-based structure for semantic enhancements.[14]
For RDFa to function correctly in XHTML and general XML documents, strict well-formedness requirements must be met, as these host languages adhere to XML syntax rules. Documents must be valid XML, with the root <html> element explicitly declaring the default namespace xmlns="http://www.w3.org/1999/xhtml" to ensure proper interpretation of elements and attributes.[14] RDFa attributes, such as rel, property, and resource, are defined in the RDFa Core namespace but integrated without requiring additional namespace prefixes unless custom vocabularies are used (e.g., xmlns:foaf="http://xmlns.com/foaf/0.1/" for FOAF terms).[14] Validation can be performed using the provided DTD or XML Schema in the specification's appendices, confirming adherence to both XHTML and RDFa constraints.[14]
Embedding RDFa in XHTML Family Markup Languages, such as XHTML 1.1, involves adding attributes directly to existing elements to express relationships and properties. For instance, in an XHTML document, a link to a person's profile might use <a href="http://example.org/alice" rel="foaf:knows" resource="http://example.org/bob">Bob</a>, generating an RDF triple indicating that Alice knows Bob using the FOAF vocabulary.[14] Similarly, metadata in the <head> section could include <meta property="dc:creator" content="Jane Doe" /> to assert authorship, with the document's DOCTYPE declared as <!DOCTYPE html PUBLIC "-//W3C//DTD XHTML+RDFa 1.0//EN" "http://www.w3.org/MarkUp/DTD/xhtml-rdfa-1.dtd"> for validation.[14] These examples demonstrate how RDFa enhances XHTML without altering its visual rendering, focusing on machine-readable annotations.
A key limitation of RDFa in XHTML and XML contexts is its reliance on XML parsers, which enforce strict syntax rules; non-XML parsers, such as those used for HTML, may fail to process the document correctly or ignore RDFa attributes altogether if served with a text/html MIME type.[14] This requirement for application/xhtml+xml serving contributed to challenges in broader adoption, especially as the XHTML 2 Working Group was discontinued by the W3C in December 2009, shifting focus away from strict XML-based evolutions of XHTML in favor of more flexible HTML standards. Despite this, RDFa support persisted through dedicated modules like XHTML+RDFa 1.1, maintaining its utility in XML environments.[26]
HTML5 Compatibility
The HTML+RDFa 1.1 specification, published as a W3C Recommendation, defines rules and guidelines for adapting RDFa Core 1.1 and RDFa Lite 1.1 to HTML5 and XHTML5, enabling the embedding of RDF triples within HTML5 documents while accommodating the language's syntax and parsing model.[3] This adaptation ensures that RDFa attributes can be processed in HTML5 serializations without requiring XML conformance, allowing broader web compatibility for semantic markup.
Unlike XML-based parsing, HTML5 employs a more lenient tokenizer that may alter element nesting or attribute placement during document loading, so RDFa processors must apply HTML5 parsing rules to inputs in HTML4, HTML5, or XHTML5 formats and correct any resulting DOM inconsistencies before generating the RDF infoset.[3] Attribute quoting in HTML5 follows the language's flexible rules, where unquoted values are permitted in many cases, but RDFa attributes like property or resource must still adhere to HTML5 syntax; additionally, @xmlns: declarations are treated case-insensitively to align with HTML5's namespace handling.[3] Void elements, such as img or br, support RDFa attributes without special restrictions, and elements like link and meta can appear in the body as flow content to facilitate metadata embedding.[3]
RDFa in HTML5 integrates seamlessly with schema.org vocabularies, where it serves as one of the three supported formats—alongside Microdata and JSON-LD—for marking up structured data to enhance search engine understanding and rich snippet generation.[27] Google's structured data guidelines explicitly support RDFa for HTML5 pages, allowing publishers to use it for features like product reviews or event details, which can improve visibility in search results through enhanced excerpts and knowledge graph integration.[28]
In digital publishing, EPUB 3.3, updated in 2025, permits RDFa attributes within its XHTML content documents—conforming to HTML5 with XML syntax—to enrich semantic structure, as specified in the EPUB core media type requirements that reference the HTML+RDFa guidelines.[29]
Benefits and Applications
Key Advantages
RDFa provides publisher independence by enabling the use of shared vocabularies identified by dereferenceable URLs, allowing data to be reused across diverse applications and websites without proprietary format lock-in. This approach treats vocabularies like web resources, where any publisher can reference and extend them, fostering interoperability in the semantic web ecosystem.[2]
A core benefit of RDFa is its ability to embed structured metadata directly within HTML documents using standard attributes, rendering the data self-contained and eliminating reliance on external files or separate data silos. This inline integration ensures that semantic information remains associated with the visible content, simplifying deployment, maintenance, and consumption by tools like search engines or data aggregators without disrupting the document's presentation.[2]
RDFa promotes modularity through support for multiple established vocabularies, such as Dublin Core for descriptive metadata, FOAF for social networks and personal profiles, and schema.org for enhancing search engine understanding of web pages. Publishers can combine these via prefix mappings or the vocab attribute, enabling flexible expression of complex domain-specific semantics within a single document while adhering to the RDF data model.[2][30]
The semantic markup in RDFa enhances machine understanding by providing explicit relationships and properties that enrich content meaning for applications like search engines and data aggregators. This aligns with broader web standards for structured data, where annotations support improved indexing and interpretation.[2]
In comparison to alternatives like Microdata and JSON-LD, RDFa delivers RDF-native expressiveness, supporting advanced features such as inheritance, internationalization via language-tagged literals, and integration of multiple ontologies, which Microdata handles less robustly due to its simpler attribute set. While Microdata prioritizes ease of adoption with a lightweight syntax, RDFa enables richer, graph-based data representations suitable for linked data applications. JSON-LD, by contrast, uses a script-based format that decouples data from markup for easier maintenance but lacks RDFa's direct embedding of semantics into HTML elements, potentially complicating alignment with visual content. As of 2024, web usage data indicates RDFa remains widely implemented but JSON-LD is growing fastest and recommended by Google.[31][2][32]
Common Use Cases
RDFa enables web publishers to embed structured metadata for various entities, such as books, events, and products, directly into HTML using the schema.org vocabulary. This approach allows content creators to annotate elements like book titles, author details, event dates and locations, or product specifications (e.g., price, availability, and reviews) with RDFa attributes, making the data machine-readable without altering the visual presentation. For instance, a publisher might use attributes like typeof="Book" and property="name" on a book description to facilitate aggregation by search engines or data crawlers. This integration with schema.org promotes semantic enrichment, as RDFa attributes map seamlessly to schema.org types and properties, supporting diverse applications from e-commerce catalogs to event calendars.[33][2]
A prominent use case is enhancing search engine optimization (SEO) through rich snippets. Major search engines, including Google and Bing, have supported RDFa for structured data markup since the early 2010s, enabling enhanced search result displays that include visual elements like star ratings for reviews, pricing for products, or schedules for events. By embedding schema.org terms via RDFa, websites can trigger these rich snippets, which provide users with more informative previews and can increase click-through rates. This capability relies on RDFa's compatibility with HTML, allowing seamless processing by search engine crawlers.[28][34]
In digital publishing, RDFa supports metadata embedding and navigation in standards like EPUB 3. EPUB 3 incorporates RDFa 1.1 Lite within its XHTML-based content documents to express publication-level metadata, such as identifiers, contributors, and structural navigation aids, directly in the readable text. This allows reading systems to extract and utilize semantic information for features like table of contents generation or accessibility enhancements, while maintaining compatibility with linked data principles. For example, RDFa can annotate navigation documents to link sections with metadata properties from Dublin Core or schema.org, improving interoperability across e-book ecosystems.[29][35]
RDFa facilitates interoperability within linked data ecosystems, particularly in public sector and cultural heritage domains. Government websites, such as those in the UK, have employed RDFa to publish structured data for resources like job vacancies and educational datasets on platforms like data.gov.uk, enabling third-party aggregation and SPARQL querying for cross-agency reuse. This markup supports the creation of linked open government data, as demonstrated in extensions to the Linked Data API using RDFa templates for datasets like Edubase. In cultural heritage, RDFa is used to embed structured metadata in digital library web pages, often leveraging schema.org for describing collections, artifacts, and relationships; analyses of global cultural heritage sites reveal RDFa instances alongside other RDF serializations to enhance discoverability and linking to broader knowledge graphs.[36]
Usage Statistics
RDFa adoption has been tracked through large-scale web crawls, such as those conducted by the Web Data Commons (WDC) project, which analyze structured data markup in billions of HTML pages from the Common Crawl corpus. In early 2013, approximately 5.61% of parsed HTML URLs contained RDFa markup, representing 169 million pages across 519,379 domains. By 2023, this figure had declined to about 2.37% of URLs (79.5 million pages across 535,458 domains), reflecting a broader shift toward JSON-LD, which rose from negligible usage in 2013 to 33.4% of URLs (1.12 billion pages) by 2023. Microdata usage, meanwhile, grew from 3.23% (97 million URLs) in 2013 to 24.5% (822 million URLs) in 2023, though both RDFa and Microdata have seen relative declines compared to the rapid rise of JSON-LD.[37][38]
The following table summarizes key adoption metrics for RDFa and competing formats based on WDC crawls:
| Year | RDFa URLs (millions) | RDFa % of Parsed HTML URLs | JSON-LD % of Parsed HTML URLs | Microdata % of Parsed HTML URLs |
|---|
| 2013 | 169 | 5.61% | ~0% | 3.23% |
| 2021 | 112 | 3.50% | 24.8% | 26.2% |
| 2022 | 91 | 2.99% | 28.8% | 26.3% |
| 2023 | 79.5 | 2.37% | 33.4% | 24.5% |
Examples
Basic XHTML RDFa 1.0 Markup
RDFa 1.0 enables the embedding of structured metadata directly into XHTML documents using a set of attributes that map to RDF triples, allowing for the expression of relationships between resources in a machine-readable format. This approach integrates seamlessly with XHTML 1.0, where the host language's strict syntax ensures well-formed markup while extending it with semantic annotations. For basic usage, RDFa 1.0 relies on attributes like xmlns for namespace declarations, about to identify the subject resource, property to specify predicates from vocabularies such as Dublin Core, and content to provide literal values when the visible text is not suitable for the metadata.
A prerequisite for valid RDFa 1.0 markup in XHTML is the use of a proper XHTML doctype, such as <!DOCTYPE html PUBLIC "-//W3C//DTD XHTML+RDFa 1.0//EN" "http://www.w3.org/MarkUp/DTD/xhtml-rdfa-1.dtd">, which declares support for RDFa extensions within the XHTML namespace. This doctype ensures compliance with XHTML 1.0 Strict rules, preventing parsing errors that could disrupt RDF extraction.
Consider a simple example of an XHTML document embedding Dublin Core metadata for a book's title, creator, and publication date. The following markup declares the Dublin Core namespace and annotates the <head> section:
xml
<!DOCTYPE html PUBLIC "-//W3C//DTD XHTML+RDFa 1.0//EN"
"http://www.w3.org/MarkUp/DTD/xhtml-rdfa-1.dtd">
<html xmlns="http://www.w3.org/1999/xhtml"
xmlns:dc="http://purl.org/dc/elements/1.1/">
<head>
<title>Sample Book</title>
<meta about="" property="dc:title" content="The Semantic Web" />
<meta about="" property="dc:creator" content="Tim Berners-Lee" />
<meta about="" property="dc:date" content="2001" />
</head>
<body>
<p>This is a sample document.</p>
</body>
</html>
<!DOCTYPE html PUBLIC "-//W3C//DTD XHTML+RDFa 1.0//EN"
"http://www.w3.org/MarkUp/DTD/xhtml-rdfa-1.dtd">
<html xmlns="http://www.w3.org/1999/xhtml"
xmlns:dc="http://purl.org/dc/elements/1.1/">
<head>
<title>Sample Book</title>
<meta about="" property="dc:title" content="The Semantic Web" />
<meta about="" property="dc:creator" content="Tim Berners-Lee" />
<meta about="" property="dc:date" content="2001" />
</head>
<body>
<p>This is a sample document.</p>
</body>
</html>
In this example, the about="" attribute (with an empty string) refers to the current document as the subject resource, while each property attribute links to a Dublin Core predicate, and content supplies the literal value for the object.
Breaking down the attribute usage step by step: First, the xmlns:dc declaration in the <html> element binds the dc prefix to the Dublin Core namespace URI, enabling shorthand for predicates like dc:title. Second, the about="" on each <meta> element establishes the document root (<html>) as the subject, creating a context for the triples. Third, property="dc:title" specifies the predicate, and since no resource is chained, it generates a literal object from the content attribute value, forming the triple <> dc:title "The Semantic Web". Similarly, property="dc:creator" yields <> dc:creator "[Tim Berners-Lee](/page/Tim_Berners-Lee)", and property="dc:date" yields <> dc:date "2001", all as literal objects in the default graph. These attributes are processed in document order, inheriting the subject context unless overridden, resulting in three independent triples without complex chaining.
When processed by an RDFa distiller, such as the W3C's RDFa 1.0 processor, this markup serializes to the following RDF/XML:
xml
<rdf:RDF xmlns:rdf="http://www.w3.org/1999/02/22-rdf-syntax-ns#"
xmlns:dc="http://purl.org/dc/elements/1.1/">
<rdf:Description rdf:about="">
<dc:title>The [Semantic Web](/page/Semantic_Web)</dc:title>
<dc:creator>[Tim Berners-Lee](/page/Tim_Berners-Lee)</dc:creator>
<dc:date>2001</dc:date>
</rdf:Description>
</rdf:RDF>
<rdf:RDF xmlns:rdf="http://www.w3.org/1999/02/22-rdf-syntax-ns#"
xmlns:dc="http://purl.org/dc/elements/1.1/">
<rdf:Description rdf:about="">
<dc:title>The [Semantic Web](/page/Semantic_Web)</dc:title>
<dc:creator>[Tim Berners-Lee](/page/Tim_Berners-Lee)</dc:creator>
<dc:date>2001</dc:date>
</rdf:Description>
</rdf:RDF>
This output represents the extracted triples in a standard RDF serialization, where the blank rdf:about="" denotes the document URI as the subject.
HTML5 RDFa 1.1 Implementation
RDFa 1.1 in HTML5 enables the embedding of RDF triples directly into HTML5 documents using a set of attributes that integrate seamlessly with the language's parsing rules, allowing for structured data expression without altering the visual rendering. This implementation adapts RDFa Core 1.1 syntax for HTML5's non-XML serialization, supporting attributes like typeof, property, resource, and vocab on any HTML5 element while ensuring conformance to HTML5's content model.[3]
A typical use case involves marking up personal information with the schema.org vocabulary, which is widely adopted for enhancing search engine understanding of web content. For example, the following HTML5 snippet defines a person entity:
html
<div vocab="https://schema.org/" typeof="Person">
<img property="image" src="janedoe.jpg" alt="Photo of Jane Doe"/>
<span property="name">Jane Doe</span>
<span property="jobTitle">[Professor](/page/Professor)</span>
<div property="address" typeof="PostalAddress">
<span property="streetAddress">20341 Whitworth Institute</span>
<span property="addressLocality">405 Whitworth</span>
<span property="addressRegion">Seattle WA 98052</span>
<span property="postalCode">98052</span>
<span property="telephone">(425) 123-4567</span>
<span property="telephone">(425) 123-4568</span>
</div>
<span property="email">[email protected]</span>
<a property="url" href="http://www.janedoe.com">janedoe.com</a>
</div>
<div vocab="https://schema.org/" typeof="Person">
<img property="image" src="janedoe.jpg" alt="Photo of Jane Doe"/>
<span property="name">Jane Doe</span>
<span property="jobTitle">[Professor](/page/Professor)</span>
<div property="address" typeof="PostalAddress">
<span property="streetAddress">20341 Whitworth Institute</span>
<span property="addressLocality">405 Whitworth</span>
<span property="addressRegion">Seattle WA 98052</span>
<span property="postalCode">98052</span>
<span property="telephone">(425) 123-4567</span>
<span property="telephone">(425) 123-4568</span>
</div>
<span property="email">[email protected]</span>
<a property="url" href="http://www.janedoe.com">janedoe.com</a>
</div>
This markup generates RDF triples such as the subject being an instance of schema:Person, with properties like schema:name "Jane Doe" and schema:jobTitle "Professor", facilitating machine-readable descriptions of individuals.[54]
RDFa attributes in HTML5 can be applied to semantic elements like <article> to annotate content sections with structured data. For instance:
html
<article vocab="https://schema.org/" typeof="Article">
<header>
<h1 property="headline">The Trouble with Bob</h1>
<time property="datePublished" datetime="2013-12-05">December 5th 2013</time>
</header>
<div property="articleBody">
<p>Blah blah blah...</p>
</div>
</article>
<article vocab="https://schema.org/" typeof="Article">
<header>
<h1 property="headline">The Trouble with Bob</h1>
<time property="datePublished" datetime="2013-12-05">December 5th 2013</time>
</header>
<div property="articleBody">
<p>Blah blah blah...</p>
</div>
</article>
Here, the <article> element serves as the container for the typed resource, with nested properties inheriting the context to form coherent triples without redundant declarations.[2]
When incorporating multiple vocabularies, the prefix attribute maps compact URIs (CURIEs) to full namespaces, enabling references across ontologies. An example declaration is prefix="schema: https://schema.org/ foaf: http://xmlns.com/foaf/0.1/", which allows usage like property="schema:name foaf:depiction" to combine schema.org for descriptive properties and FOAF for social relations in the same document, promoting interoperability.[1]
Triple extraction in HTML5 RDFa involves processing the DOM to generate RDF graphs, where inheritance plays a key role in nested structures. Parent elements establish a "current subject" via resource or about, which child elements inherit unless overridden; for example, a top-level <div resource="#me" typeof="schema:Person"> sets the subject, allowing nested <span property="schema:name">[Alice](/page/Alice)</span> to triple the same subject without repetition, thus streamlining markup for hierarchical data like a person's affiliations or events.[3]
To avoid errors in HTML5's non-XML parsing, which is tolerant but can lead to attribute misinterpretation, developers should explicitly declare the vocab or prefix at the document or element level to prevent undefined CURIEs, quote all attribute values properly to handle special characters, and use the content attribute for invisible data (e.g., <span property="schema:description" content="Machine-readable bio">Visible text</span>) rather than altering visible content. Additionally, refrain from using the deprecated @xmlns: for namespace declarations, opting instead for @prefix to ensure consistent processing across HTML5 parsers that do not enforce XML namespace rules.[1]