Knowledge Yielding Ontologies
for Transition-based Organization
  • Increase font size
  • Default font size
  • Decrease font size

WordNet-Ontology mappings


As a semantic model, the KYOTO system uses a 3-layered knowledge architecture which integrates, using formal semantic relations, three different types of resources: vocabularies, wordnets and a central ontology (Vossen and Rigau, 2010). Thus, these repositories can be developed separately but combined to form a coherent and formal model.

This ontology currently consists of around 2,000 classes divided over three layers. The top layer is based on DOLCE and OntoWordNet. The second layer are the Base Concepts (BCs) which cover at an intermediate level of abstraction all nominal and verbal WordNet synsets.  Examples of BCs are building, vehicle, animal, plant, change, move, size, weight. A third layer consists of domain classes introduced for detecting events and qualities in a particular domain.

The semantic model also provides complete mappings to the ontology for all nominal, verbal and adjectival WordNet synsets. The mappings also harmonize predicate information across different part-of-speech (POS). For instance, migratory events represented by different synsets of the verb migrate, the noun migration or the adjective migratory inherit the same ontological information corresponding to the ChangeOfResidence class.

The mapping is distributed in three tables for OntoTagging:

  • T1 contains the synset to Base Concept mapping with 114,728 records. 
  • T2 contains the synset to ontology mapping with 187,347 records.
  • T3 contains the explicit ontology with 30,000 records.
The explicit ontology is a table containing all the ontological implications. Its main purpose is to optimize the performance of the mining module over large quantities of documents. There are several advantages for the event extraction from ontotagging. First of all, the mining module can apply pattern-matching to BCs and ontological implications instead of just the words or synsets. By making the implicit ontological statements explicit, the mining module is able to find the same relations hidden in different expressions with different surface realizations, e.g.: water pollution, polluted water, pollution of water, water that is polluted directly or indirectly express the same relations. Thus, onto-tagging is a kind of off-line ontological reasoning since all inferred implications are stored inside the KAF structure. Likewise, patterns do not need to reason over concepts, which substantially improves their performance.

This generic knowledge model provides an extremely powerful basis for semantic processing in  any domain. Furthermore, through the equivalence relations of wordnets in other languages to the English WordNet, this semantic framework can also be applied to the other languages.

A detailed description of the mappings from WordNet to the KYOTO ontology and its construction is provided in deliverable D8.3 Domain extension of central ontology final.


Download the Mappings from Wordnet to the KYOTO Ontology

OntoTagging tables version 3.2 ... (including domain synsets from English Domain Wordnet in Wordnet-LMF format)


Vossen P. and Rigau G. Division of Semantic Labour in the Global WordNet Grid. Proceedings of the 5th Global WordNet Conference (GWC'10), Mumbai, India. January, 2010.






ICT-211423 - 2008 © Kyoto Consortium