信念的非修正处理方法及其自动推理研究

项目名称： 信念的非修正处理方法及其自动推理研究

项目编号： No.61272171

项目类型： 面上项目

立项/批准年度： 2013

项目学科： 自动化技术、计算机技术

项目作者： 邓安生

作者单位： 大连海事大学

项目金额： 81万元

中文摘要： 合理地处理信念之间的冲突是常识推理研究中的一个关键性问题。本项目研究信念在动态进化过程中的逻辑演算方法及其自动推理技术，对于含有冲突的信念集，通过控制推理方式但在静态上不对已有的信念进行修正，使系统推出它所能够推出的最为可靠的结论。对于子句型的信念集，建立冲突的非修正处理的形式化机制及其自动推理算法，通过对其性能和效率进行分析，给出其优化和限制策略；使用平滑代数格结构对信念进行加权，显式地描述信念的可靠程度，将其推广到基于不精确信念的推理范畴，以便于使推理能够更有效地利用启发性信息；将定性意义下的研究结果映射为一个可进行不精确推理的定量的计算模型，在其中将代数格实例化为概念格，建立其聚类约减方法；将所得到的研究结果在语义网本体不精确模型中进行验证，代替其冲突检查和矛盾删除机制。本项目研究的推理机制在容错上直观合理，有利于在推理过程中避免有用信息丢失和无用信息产生的现象。

中文关键词： 信念；非修正；逻辑；推理；算法

英文摘要： A key problem for commonsense reasoning is how to deal with the conflictions between or among beliefs. In this project we focus on the logical calculi and automated reasoning techniques in the process of dynamic evolution of beliefs. Given an inconsistent belief set, it may be able to obtain the most reliable consequences by proper control on the reasoning ways but with non static revision of the conflicting premises. Such formalism will be first provided for a clausal belief set, and its optimizations and restrictions for improvement of the performance and efficiency will also be considered. These frameworks will then be extended to the case while some reliability degrees of beliefs can be explicitly weighted by using a smooth algebraic lattice, so that the heuristic information of beliefs can be further used in the reasoning process. In order to cope with inexact reasoning,the qualitative formalism will be mapped to a computational model, in which the smooth algebraic lattice is instantiated as a concept lattice, and its cluster reduction methods will be proposed as well. Finally, our approach results will be verified in the inexact reasoning for ontology in a semantic web, rather than the present detection and deletion method of the potential inconsistencies. Our reasoning mechanism for inconsistengcy will be

英文关键词： belief；non-revision；logic；reasoning；algorithm