Semantic MediaWiki as Knowledge Graph Interface

Semantic MediaWiki positions a familiar collaborative wiki interface as a practical knowledge-graph front end—letting teams store structured entities, query them, and publish results as tables, maps, timelines, and machine-readable exports. The core pitch is that it turns a MediaWiki instance (the engine behind Wikipedia) into a flexible knowledge management system for custom domains, without forcing users to abandon the page-based workflow they already know.

At the center is an open-source extension stack that adds semantic structure to wiki pages. After installing Semantic MediaWiki, users keep the benefits of MediaWiki—versioning for every edit, collaborative editing, namespaces/categories for weak structuring, and a strong API/community—then gain the ability to attach typed properties and relationships to pages. Data can be entered directly through wiki markup, generated via templates, or collected through form-like extensions such as Page Forms. The system also supports semantic web standards, including RDF output and “triple star” support, making it easier to align a wiki-based knowledge base with knowledge-graph tooling.

The talk frames Semantic MediaWiki as a knowledge graph tool using a comprehensive knowledge-graph definition: concepts map to classes/categories, properties map to property pages (including data types), relationships are expressed through annotations in wiki text or templates, and entity descriptions live on wiki pages. It also supports schema and metadata (for example, last editor), taxonomies via categories and subcategories, and links to external identifiers and vocabularies. That structure enables FAIR-aligned practices—such as reusing controlled vocabularies and exporting data—while keeping the “instance data” anchored in the wiki’s page model.

A concrete example shows how a page about Vienna can carry typed attributes like population and coordinates, plus an external identifier (e.g., a Wikidata ID). Categories place the page into a taxonomy, while templates and forms streamline data entry. Internal querying then powers dynamic views: an Ask query can retrieve all pages in a category with a given property value (such as entities named after Sigmund Freud) and render results as tables or maps. Semantic MediaWiki offers multiple result formats, including JSON, RDF, BibTeX, and vCard, plus visualization-oriented outputs like timelines, charts, and tag clouds.

For teams that want a traditional triple store or RDF workflow, the interface can extend beyond the wiki. Semantic MediaWiki stores extra semantic data in SQL tables (often on MariaDB, SQLite, or PostgreSQL), and can optionally integrate search back ends like ElasticSearch. For SPARQL/RDF graph usage, it can connect to existing triple stores via connectors, or export RDF dumps for ingestion into external knowledge graphs. The talk also notes extensions for SPARQL querying and RDF endpoint options, while cautioning that some approaches may be less maintained.

Finally, knowledge-graph maintenance is treated as a first-class concern. The system supports “semantic gardening” through curation workflows and user rights, enabling property health checks (outdated or missing annotations, uniqueness issues, redirects/equality), and lightweight inferencing through subproperties, subcategories, and redirect-based equality. Compared with Wikibase (the technology behind Wikidata), Semantic MediaWiki is described as more flexible in its data model and more wiki-native for editing and internal querying, while both typically rely on external triple stores for SPARQL. The overall message: Semantic MediaWiki can serve as a knowledge-graph interface that keeps governance, entry, querying, and publishing inside a collaborative wiki environment—while still offering paths to RDF and triple-store ecosystems.