Uses of Ontology/KG
- Formally (machine processable) represent the knowledge of a domain
- Make the knowledge shareable (by other applications), and reusable
- The knowledge that is in your/domain expert's head is given a structure and put out
- Your intuitions, heuristics could be captured in the form of an ontology
- If the dictionaries, and vocabularies that you use have hierarchies, relationships, then an it can be represented as an ontology
- Ontology can give a good description of your data (data documentation)
- Makes the data usable
- Data can be converted to structured data (RDF)
- Data integration
- IRIs make it easy to integrate the same entities from different data sources
- Different data formats can be converted to one unified RDF format
- Ontology + KG enables automated reasoning
Uses of Reasoning
- Get complete answers
- SubClassOf(Baby Child)
- SubClassOf(Child Person)
- ClassAssertion(Baby s)
- Consistency of the knowledge base can be checked/maintained
- DisjointClasses(Vertebrates Invertebrates)
- ClassAssertion(Vertebrates xyz)
- ClassAssertion(Invertebrates xyz)
- Get explanation for the decisions/output given by the reasoner
Ontology/KG construction
- If KG is for a single/narrow domain, build an ontology first
- You would know the concepts and relations that you are interested in
- Instances (from text) can be mapped to one of the concepts
- Relations can be mapped to one of the "clean" relations in the ontology