Domain Knowledge Modelling for Educational Purposes: Methods and Tools

点击翻译按钮获取中文摘要

The idea of having a system that can teach a given subject with the effectiveness of the best human teacher (i.e., Intelligent Tutoring Systems - or ITS), has been at the fore front of many applications of artificial intelligence since its birth. These systems have tremendous potential and important economic and social benefits in many areas. They convey many challenging problems of AI. One such problem is the acquisition and representation of domain knowledge (that is, knowledge about the topic being taught), as well as reasoning about this knowledge. The main goal of this research proposal is to develop methods and tools to enhance ITSs with automated facilities capable of building, improving and maintaining an effective semantic and procedural domain knowledge model. First, I will develop new methods for refining, completing, and improving the domain ontology within an ITS. Those methods stand to contribute to the knowledge engineering aspects of ITSs, as well as other knowledge-based or knowledge management systems. The provision of tools that can facilitate domain ontology creation and maintenance, will likely produce further ITSs with ontology-based domain knowledge modules, and thereby make it easier to share domain knowledge between different ITSs. Second, I will elaborate and test effective data mining algorithms that can yield useful procedural knowledge from educational problem-solving data, particularly in ill-defined domains. Such procedural knowledge will be used to provide useful tutoring services to learners in domain where a clear task model is difficult to set up. We expect our findings to contribute to current educational data mining efforts, by offering valuable alternatives and complementary opportunities for the development of ITS in ill-defined domains. This research program is bound to produce methods and tools, that will particularly enhance the performance of ITS and e-learning software in dynamic, evolving domains, unburdened by traditional concerns for capturing semantic and procedural domain knowledge. Such results will offer Canadian e-learning providers software and service enhancement options, and help other Canadian companies to improve their knowledge management platforms.