The Top Ontology (TO) (Alonge et al., 1998) is an indepedent hierarchy of features designed for clustering, comparing and exchanging concepts across languages in the EuroWordNet Project (Vossen, 1998). Furthermore, it has been usually used as a repository of lexical semantic information. Each WordNet synset has been annotated to one or more TO feature.
In the following link, we have the annotation of WordNet 1.6 to TO (version 2.3):
The TCO consists in 63 features organized in three disjoint types of entities:
- 1stOrderEntity: physical things (image)
- 2ndOrderEntity: events, states and properties (image)
- 3rdOrderEntity: unobservable entities
Most of the subdivisions of the TO are disjoint categories: a concept cannot be both Natural and Artifact. Nevertheless, some of these inconsistences can be found when the TO features are inherited through the hyponymy hierarchy.
We can avoid the inheritance of disjoint categories including some blockage points in the hyponymy hierarchy paths. In this way, a consistent annotation of the nominal part of WorNet is obtained.
WordNet to TO Annotation Tools
We have developed a set of tools for checking the consistency of the annotation and also obtaining its expansion. For proving consistency, we check that there is no incompatiblity in the annotation of the nominal part of WordNet 1.6 to TO when using the blockage points. The expansion of the annotation can be obtained when the annotation is consistent. These tools have been implemented in Prolog and are available in the following links: [tar.gz] [zip]
We have gotten some interesting numeric conclusions from the TO annotation and the addition of the blockage points. For instance, every blockage point subsumes an average of 120.16 synsets; there are 28,123 synsets that have at least one blockage point in their hypernymy line.
This package is distributed under Attribution 3.0 Unported (CC BY 3.0) license. You can find it at http://creativecommons.org/licenses/by/3.0