Knowledge graph
A knowledge graph is a structured representation of real-world entities and the relationships between them, stored as a directed graph where nodes represen...
Networks of Facts for Intelligent Systems
A knowledge graph is a structured representation of real-world entities and the relationships between them, stored as a directed graph where nodes represent entities and edges represent relationship types. Knowledge graphs provide AI systems with a rich, queryable factual context that supports reasoning, question answering, and recommendations.
Each edge in a knowledge graph is a triple: subject, predicate, object. For example, the triple (Albert Einstein, bornIn, Ulm) states a specific fact about a specific entity. Millions of such triples form a dense network of interconnected facts. The schema defining allowed entity types and relationship types is called an ontology.
Large-scale knowledge graphs include Wikidata, which contains more than a billion statements about entities extracted from Wikipedia and structured data sources. Google's Knowledge Graph powers the information panels displayed alongside search results. The graph allows Google to answer factual questions directly — "Who is the CEO of Apple?" — by traversing relationship edges rather than retrieving web pages.
A concrete example: in an e-commerce recommendation system, a knowledge graph might link products to categories, brands, compatible accessories, and customer demographics. When a user buys a DSLR camera, the system traverses the graph to find that this product links to lenses, tripods, and camera bags, generating relevant recommendations without relying solely on collaborative filtering.
Knowledge graphs are also used in drug discovery, where they link genes, proteins, diseases, and chemical compounds. By finding unexpected paths through the graph — for example, a drug already approved for one disease sharing molecular targets with another — researchers identify repurposing opportunities.
RDF (Resource Description Framework) and SPARQL are the standard technologies for representing and querying knowledge graphs on the semantic web.
Exam tip: Know the triple format and be able to give examples. Know at least one real-world knowledge graph (Wikidata, Google Knowledge Graph) and one application domain beyond web search.