Issues: 1. Where do concepts like “Container” go in the taxonomy -- does everything go underneath Container? 2. What slots does the SME see when looking at an object? One approach is to put every conceivable abstraction above every object. Each object then inherits zillions of slots. We then manually pick off the slots which actually should be presented to the SME. Here is an alternative approach: We only link objects to an object’s standard abstractions (ie. Commonly used abstractions). Thus we say Cup isa Container, but not Cell isa Container. The SME then only sees the slots from those abstractions. However, if the SME wants to use other slots, he/she can pick them out from the slot vocabulary -- and by picking a new slot, this will automatically add the abstraction where that slot originates onto the object. So if the SME says “the cell contains cytoplasm”, this pulls in the container model ie. Cell isa Container is added to the taxonomy, and the SME now does see all the container slots. Storyboard illustration follows.Pete and John 10/20/00
Physical-ObjectContainer Cell superclasses Thing contains: permeability: portals: size: shape: 0. The initial KB
Physical-ObjectContainer Cell superclasses Thing contains: permeability: portals: size: shape: 1. The SME creates Bacterial-Cell. He/she sees only the slots inherited from the superclasses. Bacterial Cell size: shape:
Physical-ObjectContainer Cell superclasses Thing contains: permeability: portals: size: shape: 2. Now, he/she wants to provide information on a slot which isn’t listed, e.g. bacterial cells contain cytoplasm. So, he/she can open up the whole slot dictionary, and pick the needed slot (“contains”) Bacterial Cell size: shape: Which relation do you want to use? causes enables entails inhibits contains prevents... (SME selects this)
Physical-ObjectContainer Cell superclasses Thing contains: permeability: portals: size: shape: 3. Because the SME has picked a slot from the “Container” component, he/she’s indirectly appealed to the Container model. So Shaken adds the container model in with a “superclass” link. As a result, the SME now sees the “contains” slot + the other container slots. Bacterial Cell size: shape: contains: permeability: portals:
Physical-ObjectContainer Cell superclasses Thing contains: permeability: portals: size: shape: 4. The SME can then select the aspects of container which he/she cares about for the representation he/she is building. The selection controls which subsets of slots will be exposed to him/her. Bacterial Cell size: shape: Which aspects of container do you want to import? Capacity Rigidity Portals Permeability Shape Pressure...
Physical-ObjectContainer Cell superclasses Thing contains: permeability: portals: size: shape: 5. If there were several models using the “contains” slot, then Shaken would query the SME as to which model he/she had in mind when he/she used the “contains” relation. Bacterial Cell size: shape: contains: capacity: occupied-volume: permeable-to: