RatioLog - Rational Extensions of Logical Reasoning

人类的推理并不严格遵循经典逻辑的规则。对此的解释可能是不完整的知识，不正确的信念或不一致的规范。从人工智能（AI）研究的开始起，就一直强调将理性机制纳入AI推理系统。理性不能仅仅局限于认知任务，而是涉及复杂的行为以及与其他主体和（物理和空间）环境的相互作用。本项目旨在建立推理和行为的联合模型。为此，我们将结合联合收割机逻辑推理与连续系统的建模，建立在以前的工作，非单调演算，定性空间表示，和混合自动机。我们将通过各种形式的非单调方面来扩展经典逻辑推理，例如abduc- tion或可废止论证。这将不会只在理论层面上完成，但这些扩展将被纳入现有的推理系统E-KRHyper。使用E-KRHyper和自由百科全书Wikipedia回答自然语言问题的开放域问答系统LogAnswer将变成一个理性问答系统，为评估理性推理提供了一个很好的测试平台。

Human reasoning does not strictly follow the rules of classical logic. Explanations for this may be incomplete knowledge, incorrect beliefs, or inconsistent norms. From the very begin- ning of artificial intelligence (AI) research, there has been a strong emphasis on incorporating mechanisms for rationality into AI reasoning systems. Rationality cannot be restricted to cog- nitive tasks solely, but involves complex behavior and interaction with other subjects and the (physical and spatial) environment.This project aims at establishing a joint model for reasoning and behavior. For this, we will combine logical reasoning with the modeling of continuous systems by building on previous works on non-monotonic calculi, qualitative spatial representation, and hybrid automata. We will extend classical logical reasoning by various forms of non-monotonic aspects, e.g. abduc- tion or defeasible argumentation. This will not be done on a theoretical level only, but these ex- tensions will be incorporated into the existing reasoning system E-KRHyper. The open domain question answering (QA) system LogAnswer that uses E-KRHyper and the free encyclopedia Wikipedia for answering natural language questions will be turned into a system for rational question answering, which offers an excellent testbed for evaluating rational reasoning.