Cyc Mapping Methodology

Cyc is the largest (and oldest) knowledge base in existence that has been designed for machine use and reasoning. It also may have the largest number of external knowledge bases and ontologies mapped to it. As a result, its published methodologies for ontology mapping deserves distinct treatment.

This basic Cyc mapping methodology is described in a Cycorp white paper and related publications.

Importation of an Ontology

 * Create or Match terms in Cyc to terms in ontology
 * Initial mapping
 * Translate definitional assertions to CycL
 * Translate non-definitional assertions to CycL
 * Search for matching Cyc concepts
 * Merge with matching Cyc concepts
 * Refine mapping
 * Assisted refinement of information
 * Specialized tools built for importation of massive databases and taxonomies (e.g., NAIC -- North American Industry Classification System)
 * "Slurping" tools used for standard databases
 * Shell/editor scripts used for tabular data
 * These tools provide a first pass -- generating classes, relations, individuals, & mappings, and expressing relationships specified in the source.
 * Phrase parsing for (proposed) names
 *  #$isa s and  #$genl s connect to Cyc ontology
 * Second pass: generality level, suggested KE.
 * If an ontology being imported has a notion that is not in Cyc, a term for that notion is normally added to the Cyc ontology.
 * Functional terms, using  #$MeaningInSystemFn, may be created if there are no obvious direct mappings so that assertions in the external ontology can be represented.
 * Functional term definition may be used
 * All x such that y
 *  #$SubcollectionWithRelation[From|To][Type]Fn
 * As Cyc terms are created (or discovered) with the correct meanings
 * The functional term is normally replaced.
 * A  #$synonymousExternalConcept assertion is made to express the mapping.

Importation Second Pass

 * Redundancy Removal
 * Assertion subsumed by existing more general assertion
 * Generality Level Checking
 * Should assertion be more general/specific
 * Options given of term replacements
 * Intermediate Class Suggestion
 * Multiple sibling classes with similar assertions
 * Suggested Knowledge Entering