OWL-miner is a top-down refinement-based learner that searches for useful class descriptions by enumerating every possible expression and testing them for quality against the training data. It starts with a very simple expression and gradually builds a tree of more complex expressions from there, checking them as it goes.
Actually the starting class is owl:Thing, the class of all individuals, for which every other class is a subclass, In Michalski Trains, for example, one step of downward refinement would generate the candidate hypotheses Car, Count, _and _Shape _amongst others. Other steps rooted at _Car _would generate, amongst other candidates, the hypothesis Closed (because _Closed is a direct subclass of Car) and also the hypothesis Car hasload owl:Thing (i.e a Car that has a load of any class, because this is a single-step extension of Car with a property). N.B the overly-simplified Turtle notation here is not actually OWL.
若有收获，就点个赞吧
0 人点赞
