Regarding automatic inferencing
by Niels-Oliver Walkowski
on the 4. August 2010, 09:16 o'clock
As an reply to the OKF Blog entriy about machine reasoning I want to outline some personal concerns. The problem of automatic inference is not the lack of context which could be compensated as you have demonstrated. It is the fact that context from the perspective of the rdf statement or an inference machine is in many cases indescribable. Not because there is an ontology missing but because there are many possible contexts imaginable. A very simple phenomena of this problem is the ambiguity of many words. The more abstract the concept the more it could be integrated in different contexts. The next problem is that Ontologies like OpenSync represents a particular perspective about the world which may not fit for all people in their own situations. This problem my not be so dramatic with simple things like sheeps and lambs but it becomes a problem with more theoretical concepts. One dream of automatic interference apostles was always to bring it into science and to generate new knowledge with it. In my opinion this is impossible because of the problems mentioned. Interference is build up upon reasoning and reasoning is a deterministic thing: you can only get out what you put in. This is not the way scientific innovations work.
