Semantics and Proof Procedure for Abductive Logic Programming

一般逻辑程序的语义可以用二值稳定模型和三值稳定模型来定义，而一般逻辑程序的完成过程采用SLDNF归结（SLD归结，否定为失败）。当SLDNF解析被很好地组织为溯因证明过程（Eshghi和Kowalski过程）的基础时，首先检查该过程对于2值稳定模型是否是合理的。本文研究了一个三值逻辑的溯因框架，它是Eshghi和Kowalski框架的一个推广，在应用否定作为失败时，我们通常采用一个安全的规则：*（]SY. [)A如果 * A失败，则成功，并且 *（]SY.反转左半括号。[)A如果 * A成功，则R1失败，其中A是基原子。向ge ...更多信息 t非地面可展性，我们宁愿采取谢泼德森规则（非安全规则）：*（]SY.反向左半括号。[）Arheta成功，如果 * Arheta失败了一些替代rheta，和 *（]SY.反向左半括号。[)A对于溯因证明过程，如果 * A通过空替换成功，则证明失败。本文证明了一个改进的溯因证明过程在一般的三值稳定模型下的可靠性，该模型定义在一个包含变量的域上.作为另一方面，我们在三值逻辑中有一个溯因框架，其中一个完整性约束被组织起来，使得它可以导出一个广义的溯因证明过程，这涉及到一个调整，以处理任何命题既不是一个溯因，也不是来自理论和溯因。本文证明了完整性约束与三值稳定模型有密切的关系，并将其分为三个部分：（1）逻辑程序的语义。(2)SLDNF分辨率的改进。(3)外展程序和外展框架的概括。少

Semantics of general logic programs may be defined by both 2-valued stable and 3-valued stable models, while SLDNF resolution (SLD resolution with negation as failure) is adopted as a procedure for the completion of general logic programs. When SLDNF resolution was well-organized as a basis of abductive proof procedure (Eshghi and Kowalski procedure), it was firstly examined whether the procedure is sound with respect to 2-valued stable model. However, it has been proved to be sound with respect to-3-valued stable model (Dung, 1991).This research project is concerned with a generalization of negation as failure to be made use of for an abductive proof procedure, and an abduction framework in 3-valued logtic, a generalization of Eshghi and Kowalski framework.When applying negation as failure, we usually take a safe ruls :*(]SY.reverse left half-bracket.[)A suceeds if * A finitely fails, and*(]SY.reverse left half-bracket.[)A finitely fails if * A succeeds,where A is a ground atom. To ge … More t non-ground abducibles, we rather take Shepherdson's rule (a non-safe rule) :*(]SY.reverse left half-bracket.[)Arheta succeeds if * Arheta finitely fails for some substitution rheta, and*(]SY.reverse left half-bracket.[)A finitely fails if * A succeeds with the empty substitution,for an abductive proof procedure. Soundness of a refined abductive proof procedure is proven with respect to a general 3-valued stable model, which is defined over a domain containing variables.As another aspect, we have an abduction framework in 3-valued logic, where an integrity constraint is organized so that it may induce a generalized abductive proof procedure, which involves an adjustment to deal with any proposition to be neither an abducible nor derived from the theory and the abducibles. The integrity constraint is proved to be in a close relationship with a 3-valued stable model.This reoport comprises these two aspects in three parts :(1) Semantics for logic programs.(2) Refinements of SLDNF resolution.(3) Generalizations of abductive procedures and abduction framework. Less