Knowledge Concepts
What is a knowledge graph? A clear explanation of the technology, how Google's knowledge graph works, and what personal knowledge graphs in tools like Obsidian actually do.
A knowledge graph is a structured network of entities (people, places, concepts, things) and the relationships between them — representing knowledge as nodes connected by labeled edges, enabling computers and humans to navigate relationships between concepts rather than just searching for keyword matches.
At the technology level, Google's Knowledge Graph powers the information panels you see when you search for a person, place, or concept. At the personal level, the graph views in Obsidian and Roam Research are simplified knowledge graphs of your linked notes.
The term "knowledge graph" was popularized by Google in a 2012 blog post announcing the Google Knowledge Graph — a structured database of over 500 million entities and 3.5 billion facts connecting them. The system enabled Google to answer factual questions directly rather than just returning links.
But the concept predates Google significantly:
Semantic Web (1999): Tim Berners-Lee proposed extending the web with structured data that machines could understand — the foundational vision that knowledge graphs implement.
Resource Description Framework (RDF, 1999): W3C standard for representing relationships as subject-predicate-object triples (e.g., "Einstein — born in — Ulm"). RDF is the technical foundation of many knowledge graphs.
Wordnet (1985): Princeton's lexical database organizing English words into semantic networks — an early human-built knowledge graph.
Knowledge representation in AI (1970s–80s): Semantic networks in early AI systems represented entities and relationships as graph structures.
The modern PKM use of "knowledge graph" (Obsidian's graph view, etc.) borrows the visual metaphor of nodes and edges but is structurally simpler — it's a graph of notes, not a formal ontological knowledge representation.
At the technical level, a knowledge graph represents information as triples:
Subject → Predicate → Object
[Einstein] → [born in] → [Ulm]
[Einstein] → [developed] → [Special relativity]
[Special relativity] → [part of] → [Physics]
[Physics] → [field of study at] → [MIT]
Each triple is a fact. A knowledge graph is a collection of millions of such triples, stored in a graph database (Neo4j, Amazon Neptune, etc.) and queryable via SPARQL or Cypher (graph query languages).
The key advantage over relational databases: you can traverse relationships without knowing the schema in advance. "Give me all the entities two hops away from Einstein" is a natural graph query; in SQL, it requires complex JOINs that must be designed upfront.
Enterprise knowledge graphs: Structured, formal knowledge bases used by organizations. Google's Knowledge Graph, Facebook's social graph, LinkedIn's professional graph. Also internal corporate knowledge graphs that connect employees, projects, documents, and relationships.
Domain-specific knowledge graphs: Built for a specific field. Biomedical knowledge graphs (connecting genes, diseases, drugs, and clinical trials), financial knowledge graphs, legal knowledge graphs. Wikidata is a community-built, publicly accessible general-purpose knowledge graph.
Personal knowledge graphs: The informal version — your linked notes in Obsidian, Roam, or Logseq. The graph view shows notes as nodes and links as edges. Not formally a knowledge graph in the technical RDF sense, but functionally similar: a visual representation of relationships between pieces of knowledge.
When PKM practitioners talk about their "knowledge graph," they usually mean the visual graph view of their linked notes system, not a formal semantic graph.
In Obsidian's graph view:
[[wikilinks]] between notesThis is valuable for:
The limitation: PKM "knowledge graphs" aren't semantically labeled — the edge just means "these two notes link to each other." A formal knowledge graph would label that edge: "this note supports that note" or "this concept contradicts that concept."
Search Google for "Albert Einstein."
The Knowledge Panel appears on the right (desktop) or at the top (mobile) with:
All of this comes from the Knowledge Graph — structured entity data retrieved and displayed without needing to rank and display web pages for every field.
Behind the scenes: the Knowledge Graph contains a node for "Albert Einstein" with edges to nodes for "Ulm" (birthplace), "Elsa Einstein" (spouse), "Special relativity" (notable work), and hundreds of other entities. When you search "Einstein," Google retrieves this structured subgraph and renders the panel.
A researcher using Obsidian for 18 months has 800 notes. She opens the graph view:
Insight: behavioral economics and SaaS pricing are the two best-developed areas of her knowledge. The supply chain notes are orphaned — she should either connect them to other notes or acknowledge that she hasn't developed this area.
She clicks the behavioral economics cluster and navigates note-to-note, discovering a connection she'd missed: a note on loss aversion links to a note on subscription cancellation rates that she should connect to her SaaS pricing work.
"Knowledge graphs are only for large tech companies." Wikidata (public, free), Apache Jena (open-source), and various cloud services (Amazon Neptune, Neo4j) make knowledge graph technology accessible to developers at any scale. Personal knowledge graphs in Obsidian need no technical knowledge at all.
"The graph view in Obsidian is a knowledge graph." Loosely, yes — it's a visual graph of linked knowledge. But it's not a knowledge graph in the formal sense: there's no ontology, no labeled predicates, no structured data. It's a visual metaphor, not a technical implementation.
"Knowledge graphs require AI to build." Early knowledge graphs were built by human curators (Freebase, Wikidata). Modern knowledge graphs use NLP and ML to extract entities and relationships from text automatically — but the technology is separate from the data model. You can build a perfectly valid knowledge graph by hand.
WebSnips's Connections graph is a knowledge graph for your personal research library:
This is more sophisticated than Obsidian's graph view (where you manually create links) because connections are discovered automatically from content. For knowledge workers whose primary research source is web content, WebSnips's Connections graph provides the relationship visibility that Obsidian's graph provides for note-takers.
What's the difference between a knowledge graph and a database? A relational database stores structured data in tables with predefined schemas and relationships. A knowledge graph stores entities and relationships flexibly, without requiring a predefined schema, and is optimized for traversing relationship paths rather than row-level queries. Knowledge graphs handle "everything related to X" queries efficiently; relational databases handle "all rows where field = value" efficiently.
How does Google's Knowledge Graph get updated? Google's Knowledge Graph is updated continuously via a combination of automated extraction from web content, structured data (schema.org markup on websites), partnerships with authoritative data sources (Wikipedia, Freebase's successor Wikidata), and direct submissions. No single update mechanism — it's a constantly evolving combination.
Can I build my own knowledge graph? Yes — at various levels of sophistication. For a personal PKM knowledge graph: just create a linked notes system in Obsidian. For a domain-specific technical knowledge graph: Neo4j (graph database) + a SPARQL or Cypher interface + NLP extraction tools. For a public contribution: Wikidata accepts community edits.
What's Wikidata? Wikidata is a free, collaborative knowledge graph maintained by the Wikimedia Foundation — structurally similar to Wikipedia but containing structured data (entities + relationships) rather than prose articles. It has over 100 million data items and is publicly accessible via API. It's the largest open-access knowledge graph available.
A knowledge graph is a structured network of entities and their relationships, enabling navigation by connection rather than keyword search — whether at Google's scale (billions of entity-fact triples) or at personal scale (hundreds of linked notes in Obsidian).
For knowledge workers, the personal knowledge graph is the destination of a good PKM practice: a web of connected ideas, notes, and sources that can be navigated, not just searched.
More WebSnips articles that pair well with this topic.
What is citation management? A clear explanation for researchers and academics — how citation managers work, the best tools, and how to avoid the most common mistakes.
What is tacit knowledge? A clear explanation for team leads and ops people — what tacit knowledge is, why it's the hardest knowledge to transfer, and practical methods to make it explicit.
What is a commonplace book? A clear explanation of the centuries-old knowledge collection practice, famous examples, and how to build a modern digital version for writers and researchers.