Taxonomical hierarchy of canonicalized relations from multiple Knowledge Bases

Paper: Taxonomical hierarchy of canonicalized relations from multiple Knowledge Bases
Website: Relation Hierarchy

Abstract

This work addresses two important questions pertinent to Relation Extraction (RE). First, what are all possible relations that could exist between any two given entity types? Second, how do we define an unambiguous taxonomical (is-a) hierarchy among the identified relations? To address the first question, we use three resources Wikipedia Infobox, Wikidata, and DBpedia. This study focuses on relations between person, _organization and location entity types. We exploit Wikidata and DBpedia in a data-driven manner, and Wikipedia Infobox templates manually to generate lists of relations. Further, to address the second question, we canonicalize, filter, and combine the identified relations from the three resources to construct a taxonomical hierarchy. This hierarchy contains 623 canonical relations with highest contribution from Wikipedia Infobox followed by DBpedia and Wikidata. The generated relation list subsumes an average of 85% of relations from RE datasets when entity types are restricted.

Resources

• Hierarchy Visualtion (Check Website): We have provided different hierarchies based on the source knowledge base(or combination of Knowledge bases).
  • Dbpedia, Wikipedia, and Wikidata: individual hierarchies for person, location, and organization along with the complete hierarchy.
  • Dbpedia Wikidata: A joint hierarchy for the relations collected from both Dbpedia and Wikidata. Hierarchy can be visualised either radially or in spiral radial tree.
  • Dbpedia Wikidata Infobox: A joint hierarchy for the relations collected from both Dbpedia, Wikidata and Wikipedia Infobox templates. Hierarchy can be visualised either radially or in spiral radial tree.
• relation.csv: This csv file contains final list of 623 relations. Every line is of the form, [source-entity-type.target-entiy-type.[...ancestor-relations...].relation]

per.per.student

In above example, source entity type is person, target entity type is person, and relation is student.

per.per.student.doctoralStudent

In the above example, relation of interest is doctoralStudent and it has one ancestor relation, student.

• relation_id.csv: This csv file contains all the relation following hierarchical visualization (Radial tree/Tidy tree). Every line contains the relation path from root node to the relation itself and unique relation id.

rel.per.per-org.associatedWith.institution.professorshipAt,R10300210

In the above example, rel is the root node (common for all), professorshipAt is the relation of interst, and per (implies, person specific relation), per-org (implies, source (person) and target (organization) entity types ), associatedWith, and institution are intermdiary nodes/ancestor relations (genralisation). And R10300210 is the unique id assigned to the relation path.