Knowledge Discovery and modeling in Intelligent Tutoring Systems

知识获取一直是知识基础系统（KBS）的主要瓶颈。这对于智能辅导系统（ITS）来说尤其如此，其知识库包括领域本体和问题解决知识。为了使信息技术系统在知识迅速发展的领域蓬勃发展，例如商业和工业环境，必须开发新的易于使用的工具来获取和维护知识。第一个知识创建问题的根源在于知识库通常必须从头开始手动创建。然而，关键信息已经存在，隐藏在社区和组织内的大量现有文档和数据库中。领域专家需要将其明确并提供相关知识给人工导师，以便它能够引导学习者通过能力和解决问题的技能发展。然而，在一些领域，解决问题的知识是不容易预先确定的，它是不断发展的。为了找到更有效的解决这些问题的方法，我建议研究知识发现（KD）技术来在ITS中构建领域智能。首先，该项目将产生一些KD技术，这些技术可以通过挖掘将在学习发生的上下文中流动的文档来支持动态领域中的新领域知识发现和本体进化。其次，我感兴趣的是从用户的行为轨迹中自动学习任务模型，以提取新的过程（正确或不正确），新的问题空间和新的问题解决策略。我们知道，为了产生有效的问题空间或任务模型，以支持有价值的辅导服务（如模型和知识跟踪、指导、错误检测和计划识别），认知任务分析和规则编写过程是多么耗时。当前的创作系统无法简化这些过程。这个项目提供了一个很好的机会来克服ITS中知识获取的两个主要障碍。

Knowledge acquisition has always been the major bottleneck for knowledge-based systems (KBS). This is especially true for Intelligent Tutoring Systems (ITS) in which knowledge base includes domain ontology and problem solving knowledge.  In order for ITSs to thrive in fields where knowledge evolves rapidly, such as business and industrial settings, new, easy-to-use tools must be developed for knowledge acquisition and maintenance. The first knowledge creation problem finds its source in the fact that the knowledge base usually has to be created manually, from scratch. However, crucial information already exists, hidden in legions of existing documents and databases within communities and organizations. Domain expert are required to make it explicit and provide relevant knowledge to the artificial tutor so that it will be able to guide the learner through competence and problem solving skills development. However, in several domains, problem solving knowledge is not easy to predetermine and it evolves.To find more efficient solutions to these problems, I propose to investigate Knowledge Discovery (KD) techniques to build domain intelligence within an ITS.  First, this project will produce some KD techniques that can support new domain knowledge discovery and ontology evolution in dynamic domains by mining documents flowing in the context where learning will take place. Second, I am interested in the automatic learning of task models from users' action traces to extract new procedures (correct or incorrect), new problem spaces and new problem solving strategies. We know how time-consuming are both the cognitive task analysis and rules authoring process aimed at producing effective problem spaces or task models to support valuable tutoring services such as model and knowledge tracing, coaching, error detection and plan recognition. Current authoring systems fail to simplify these processes. This project offers a great opportunity to overcome those two main stumbling blocks of knowledge acquisition in ITS.