Interoperability between ontologies

Ontologies can differ in several axes. The following are from Approaches to Legal Ontologies:

Topology

Fernández-Barrera and Sartor (2010) divide ontologies by topology. Rooted tree is a strict is-a taxonomy, whereas operational family has a more diverse structure, with loops.

In practice, most ontologies allow both: is-a hierarchy as a backbone, and other relations in addition.

Granularity

An ontology can be domain-specific, meant to cover a domain in fine detail. Contrast with upper ontologies, which are more general and have a small number of concepts. Core ontology is used for a general ontology within a domain: an “upper ontology” for just legal domain is better described as a legal core ontology. For more details, see Top ontology and domain ontologies#.

Formality

A lightweight (or language-oriented) ontology is not much more than a collection of concepts: little hierarchy, few axioms. Technically, a completely flat lexicon is a lightweight ontology.

In contrast, heavyweight (more formal or highly axiomatized) ontologies have more structure. Biasiotti and Tiscornia (2010) describe as follows: “Formal ontologies are composed of a relatively small set of concepts, defined by a high number of constraints which encode the relations between individuals of classes through cardinality restrictions, property range and domain, disjointness, transitive and symmetric properties.”

Building methods

Ontology learning from texts is bottom-up1. Find a word in a text, relate it to other words2 and put it in an ontology. The resulting ontology has usually a high level of detail (see “Granularity”), but not a lot of structure (see “Formality”).

In contrast, top-down starts from conceptual models. Take a generic concept, and build a structure by specialization. The resulting ontology is more structured–it’s more likely that an ontology built top-down is heavyweight and upper, but this does not have to be the case. It’s completely possible to build a lightweight, domain ontology by top-down method.

Combination of the two is called middle-out: start from the middle, generalize upwards and specialize downwards. See El Ghosh et al. (2016) for more info.

See also Conceptual ontologies and lexical ontologies#

  1. Biasiotti and Tiscornia (2010) argue that legal ontologies need to be built bottom-up: “Since legal domain is strictly dependent on its own textual nature, a methodology for ontology construction must privilege a bottom–up approaches, based on a solid theoretical model.”

    ↩︎

  2. Herbelot (2011) on ontology extraction: “— a subfield of natural language processing which, put simply, specialises in producing lists. […] Well-loved ontology extraction tasks include the retrieval of Oscar nominees, chemical reactions and dead presidents. In this kind of research, the machine is asked, for instance, to produce a list of things that are ‘like lorries’ and is expected to duly return (given the current state of the art): truck car motorcycle plane engine hamster. Because lorries have wheels and hamsters have too.”

    ↩︎
logo