Record Linking is an important aspect of any knowledge graph build process that involves linking records to entities. An example of this could include linking a corporation’s Bloomberg profile and all corresponding facts to their entity within a Knowledge Graph. For each new record, with sources from one or more origins, record linking must occur before facts are associated with an entity.
Record linking at scale leads to a variety of technical challenges including overmerging, undermerging, and issues with entity resolution. Diffbot employs sophisticated machine learning systems to aid in record linking for the billions of entities and trillions of facts in our Knowledge Graph™.