Machines that can 'think,' generally have two parts. One part does the reasoning using rules and logic.
Another part contains a formal description of some part of the world that the reasoning system refers to. Often, the reasoning half is called an inference engine, and the lookup part an 'ontology.' An ontology uses its own logic to define things. So if the inference engine wants to reason about women, the ontology has some definitions that can be used to support the inferencing.
Ontologies are designed for specific reasoning tasks in old fashioned AI systems. You couldn't know (have ontology entries for) everything about women, but if you intend to just reason about kinship relations, then the ontology will contain logical statements that tell you that women are humans, distinct in some ways from men. Girls are young women. Women can be mothers and wives, and when a mother, she might have daughters and so on.
All the initial ontology work was sponsored by DARPA. Most of the energy in the ontology community now is focused on the so-called SemanticWeb .
In theory, an ontology can be created and once done is done forever, assuming we are reasoning about the real world. (This would be like the genome project, which was a huge undertaking, but is now for all intents done.) Then we'd have a sort of structured encyclopedia of the entire world and if an expert system wanted to know anything (which it constantly does), it would just look it up.
Now as it happens, this focus on the Semantic Web is something of a disaster. The Internet was well enough designed, but the Web is an accident of simplicity and opportunistic businesses. It is a mess. The Web is built around viewing documents. Google has made it easier to search web pages than before (before Google) but has reinforced the paradigm.
The semantics of documents are quite different than those of general reasoning. We'd be much better off starting over, which is pretty much what we propose. On the other hand, Web display and control technologies are okay, and much of the behind the scenes web ontology work is good enough for us to use them. It is a matter of not reinventing the wheel, but having to use the horse and carriage it is attached to.