CRII: CHS: Fast Planning Using Computational Models of Narrative

CRII：CHS：使用叙事计算模型进行快速规划

• 批准号: 1464127
• 负责人: Stephen Ware
• 金额: $ 13.84万
• 依托单位: University of New Orleans
• 依托单位国家: 美国
• 项目类别: Continuing Grant
• 财政年份: 2015
• 资助国家: 美国
• 起止时间: 2015-05-15 至 2019-04-30
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=1464127&HistoricalAwards=false
• 关键词: CRII; CHS; Fast; Planning; Using

The objective of this research is to develop a plan-based computational model of narrative, and to use that model to automatically structure interactive experiences. When humans evaluate the importance of a narrative event, they consider not just what has happened but also events which did not happen or might have happened instead. This project will design, implement, and test a computational model of this kind of hypothetical thinking, which narrative theorists call possible worlds reasoning. This model will be used in a fast narrative planning algorithm and put to use in an existing interactive virtual environment. The results will improve our understanding of how people reason hypothetically during narrative comprehension and improve the capabilities of interactive narrative technologies to teach, train, and treat their audiences.Specifically, this research will extend previous work on hybrid planning algorithms which combine the rich knowledge representation of Partial Order Causal Link (POCL) plans with fast forward-chaining state space heuristic search. By augmenting the search space of these planners with accessibility relations the tree can be used not just as a data structure for search but also a network of possible worlds and the relationships between them. This computational model of hypothetical reasoning in narrative will be evaluated using an empirical cognitive question answering framework called QUEST. The improved algorithm, which will leverage the state-of-the-art Fast-Downward heuristic, will be evaluated in an existing interactive narrative virtual environment called The Best Laid Plans.

这项研究的目标是开发一种基于计划的叙事计算模型，并使用该模型自动构建交互体验。 当人类评估叙事事件的重要性时，他们不仅考虑已经发生的事情，而且还考虑未发生或可能发生的事件。 该项目将设计、实现和测试这种假设思维的计算模型，叙事理论家将其称为可能世界推理。 该模型将用于快速叙事规划算法，并在现有的交互式虚拟环境中使用。 研究结果将提高我们对人们在叙事理解过程中如何进行假设推理的理解，并提高交互式叙事技术教授、训练和对待观众的能力。具体来说，这项研究将扩展之前关于混合规划算法的工作，该算法将偏序因果链接（POCL）计划的丰富知识表示与快速前向链接状态空间启发式搜索相结合。 通过使用可访问性关系来增强这些规划器的搜索空间，树不仅可以用作搜索的数据结构，还可以用作可能世界及其之间关系的网络。 这种叙事中假设推理的计算模型将使用名为 QUEST 的经验认知问答框架进行评估。 改进后的算法将利用最先进的快速向下启发式算法，并将在名为“最佳计划”的现有交互式叙事虚拟环境中进行评估。