The availability of inference services in the Semantic Web context is fundamental for performing several tasks such as the consistency check of an ontology, the construction of a concept taxonomy, the concept retrieval etc.
Currently, the main approach used for performing inferences is deductive reasoning. In traditional Aristotelian logic, deductive reasoning is defined as the inference in which the (logically derived) conclusion is of no greater generality than the premises. Other logic theories define deductive reasoning as the inference in which the conclusion is just as certain as the premises. The conclusion of a deductive inference is necessitated by the premises: the premises cannot be true while the conclusion is false. Such characteristics of deductive reasoning are the reason of its usage in the SW. Indeed computing class hierarchy as well as checking ontology consistency require certain and correct results and do not need of high general conclusions with respect to the premises.
Conversely, tasks such as ontology learning, ontology population by assertions, ontology evaluation, ontology mapping and alignment require inferences that are able to return higher general conclusions with respect to the premises. To this end, inductive learning methods, based on inductive reasoning, could be effectively used. Indeed, inductive reasoning generates conclusions that are of greater generality than the premises, even if, differently from the deductive reasoning, such conclusions have less certainty than the premises. Specifically, in contrast to the deduction, the starting premises of the induction are specific (typically facts or examples) rather than general axioms. The goal of the inference is to formulate plausible general assertions explaining the given facts and that are able to predict new facts. Namely, inductive reasoning attempt to derive a complete and correct description of a given phenomenon or part of it.
It is important to mention that, of the two aspects of inductive inference: the generation of plausible hypothesis and their validation (the establishment of their truth status), only the first one is of primary interest to inductive learning research, because it is assumed that the generated hypothesis are judged by human experts and tested by known methods of deductive inference and statistics.