A node is a graph element representing an entity such as a term, person, place, thing, or concept. Edges connect nodes and represent relationships; following those edges can reveal inferred facts. section 5.1
In a semantic knowledge graph, a node is materialized by executing a query and treating its matching documents as a set. A simple term, an intersection such as “software engineer,” and a union such as “software engineer” OR “java” therefore become different document sets. Edges use the same representation by selecting documents containing both nodes. section 5.4.7
This representation supports several graph operations:
- Represent entities explicitly in a graph. section 5.1
- Materialize nodes dynamically from search queries over an index. section 5.1 section 5.4.7
- Use a relationship node to approximate an edge type that is more specific than generic relatedness. section 5.4.7
Finding who is “in love with” Jean Grey
Start with the node “jean grey,” add the relationship node “in love with,” and traverse the combined context. The resulting related terms include jean, grey, summers, cyclops, xavier, wolverine, and magneto, with traversal scores ranking their associations. section 5.4.7
