Difference (from revision 39 to 40)
Added: 82a83,109
|
Name: Specify Axioms via Examples / DB's {nid 2DN1} {{{ Motivation: I am what an ontologist / knowledge engineer would call a "domain specialist" or a "subject matter expert." I have extensive experience in my field and want to formalize my knowledge. I don't have the time nor desire to become an expert in logic, but I want to be able to express my work in a machine readable format that is shareable with others in my field and might possibly ease interface with those in fields peripherally related to mine. {nid 2DMW} Goal: Provide a mechanism for a SME to formalize intuitions. Actors: User + Axiom Generation Tool Triggers: Desire for formal axioms Preconditions: An OOR instance with at least First Order expressivity, organized by modules (i.e. COLORE) {nid 2DMX} Summary: Using the inbuilt advantages afforded with a formal language at least as expressive as first order logic, we can communicate with the SME using examples (tarski-models) only, and navigate the repository to find the best set of axioms which correspond to their intuition. {nid 2DMY} Base Course: 0) The user logs into COLORE. (it only works on COLORE atm) 1) The user names the relation(s) she wishes to formalize 2) User provides at least one example of their relation in "action" - essentially a Tarski style model. The model can be represented visually, or inputted as a plain text (see referenced papers for more on this). 3) ODT searches the ontologies in the repository (COLORE) to find "Core-Hierarchies" and bounds in each hierarchy that match the user's intuition. 4) For each Core Hierarchy, ODT presents a Tarski style model in the same representation that the user inputted. 5) The user decides if this example corresponds to their intuition. 6) repeat 5-6 until search space exhausted 7) Present user with axioms for their intuition. {nid 2DMZ} Ref: http://stl.mie.utoronto.ca/publications/design-repository.pdf (Chapter 4) https://tspace.library.utoronto.ca/bitstream/1807/17512/1/Hashemi_Ali_200906_MASc_thesis.pdf {nid 2DN0} }}} |