MODELS FORMAL EXPLICACÓN IN LOGIC AND ARTIFICIAL INTELLIGENCEAuthor:
SOLER TOSCANO FERNANDO.
Year: 2005.
University:
SEVILLA.
Place of defense: FACULTAD DE FILOSOFÍA.
Place of preparation: FACULTAD DE FILOSOFÍA.
Summary: The abduction is the kind of reasoning that, given a theory T (usually a set of formulas language with which he works) and an observation O (usually a formula), such that neither O nor his denial is a logical consequence of T it is seeking an explanation E (a formula) that So logical consequence of the union of Ty E. The abductivo reasoning, which can not be reduced to deduct poses interesting problems that go beyond the logical, as the distinction between the concepts of "abduction as a product" and "abduction as a process" or the various stages of "generation" and "selection" of the explanatory hypothesis. The reasoning abductivo finds applications in many disciplines, from Artificial Intelligence to Lingüstica through Cognitive Sciences and Philosophy of Science. In this paper we propose an estimate abductivo, we call delta-resolución, dual-resolution classical (Robinson, 1965) that has the virtue of generating hypotheses "so direct." In addition, integrates the concepts of abdución as "product" and "process" as well as the stages of "generation" and "selection". Likewise, it points out that the logical extension to non-classical is more promising in this as in other approaches. After comparing the efficiency with the system Aliseda (1997), which employs semantic tables to begin searching abductiva shows that our system pays considerably better, especially in the most complex problems, because the heuristics that remains is computationally much in most advantageous problems abductivos more common in Programming Logic and Knowledge Representation. Here, implementations are provided in SWI-Prolog of algorithms defined. Finally, we propose an extension to first-order logic which shows that although the abduction on this logic is usually unspeakable, it is possible to solve interesting problems in different domains. In addition, it leaves the door open to adaptation of partial solutions to the decision calculating delta-resolución.
