Knowledge in the Loop: Semantics Representation for Multimodal Simulative Environments
Knowledge in the Loop: Semantics Representation for Multimodal Simulative Environments
复制标题
循环中的知识：多模态模拟环境的语义表示
DOI：
10.1007/11536482_3
复制
发表时间：
2005
期刊：
影响因子：
--
通讯作者：
I. Wachsmuth
中科院分区：
文献类型：
--
作者：
Marc Erich Latoschik;P. Biermann;I. Wachsmuth
本文介绍了基于知识的技术集成到模拟虚拟现实（VR）的应用。该方法的动机是使用多模态虚拟建设作为一个例子域。提出了一种抽象的知识表示层（KRL），它具有足够的表达能力来定义各种仿真任务所需的所有数据，并为人工智能（AI）表示的集成提供了基本的形式主义。KRL支持两种不同的实现方法。第一种方法是利用并行处理将外部KRL格式转换为不同目标系统的表示格式。第二种方法使用功能可扩展的语义网络实现KRL。语义网库是为真实的实时仿真系统量身定制的，其中它互连所需的仿真模块并建立对仿真循环内知识表示的访问。KRL促进了一种新的对象模型，称为语义实体的模拟对象，它提供了一个统一的访问KRL，并允许广泛的系统模块化。KRL方法在两个模拟区域进行了演示。首先，提出了一种广义场景图表示，它引入了几何节点相互关系的抽象定义和实现。它支持场景和应用程序结构，这些结构不能使用常见的场景层次结构或字段路由概念来表示。其次，KRL的表现力在多模态交互的设计中得到了证明。在这里，KRL定义了在多模态用户话语的语义分析期间特别需要的知识。
This article describes the integration of knowledge based techniques into simulative Virtual Reality (VR) applications. The approach is motivated using multimodal Virtual Construction as an example domain. An abstract Knowledge Representation Layer (KRL) is proposed which is expressive enough to define all necessary data for diverse simulation tasks and which additionally provides a base formalism for the integration of Artificial Intelligence (AI) representations. The KRL supports two different implementation methods. The first method uses XSLT processing to transform the external KRL format into the representation formats of the diverse target systems. The second method implements the KRL using a functionally extendable semantic network. The semantic net library is tailored for real time simulation systems where it interconnects the required simulation modules and establishes access to the knowledge representations inside the simulation loop. The KRL promotes a novel object model for simulated objects called Semantic Entities which provides a uniform access to the KRL and which allows extensive system modularization. The KRL approach is demonstrated in two simulation areas. First, a generalized scene graph representation is presented which introduces an abstract definition and implementation of geometric node interrelations. It supports scene and application structures which can not be expressed using common scene hierarchies or field route concepts. Second, the KRL's expressiveness is demonstrated in the design of multimodal interactions. Here, the KRL defines the knowledge particularly required during the semantic analysis of multimodal user utterances.