文章基本信息

标题：Web-based learning through mixed-initiative interactions: Design and implementation
本地全文：下载
作者：Nantha Kumar Subramaniam
期刊名称：Asian Association of Open Universities Journal
印刷版ISSN：2414-6994
出版年度：2013
卷号：8
期号：2
页码：19-31
DOI：10.1108/AAOUJ-08-02-2013-B002
摘要：Mixed-initiative interaction is a naturally-occurring feature of human-human interactions. It is characterised by turn-taking, frequent change of focus, agenda and control among the "speakers". This human-based mixed-initiative interaction can be implemented through mixed-initiative systems. This is a popular approach to building intelligent systems that can collaborate naturally and effectively with people. Mixed-initiative systems exhibit various degrees of involvement with regards to the initiatives taken by the user or the system. In any discourse, the initiative may be shared between either, a learner and a system agent, or between two independent system agents. Both the parties in question establish and maintain a common goal and context, and proceed with an interaction mechanism involving initiative taking that optimises their progress towards the goal. However, the application of mixed-initiative interaction in web-based learning is very much limited. This paper discusses the design and implementation of a web-based learning system through mixedinitiative system known as JavaLearn. JavaLearn allows the interaction between the system (in the form of a software agent) and the individual learner. Here, the system supports the learning through a problem solving activity by demanding active learning behaviour from the learner with minimal natural language understanding by the agent and embodies the application-dependent aspects of the discourse. It guides the learner to solve the problem by giving adaptive advice, hints and engages the learner in the real time interaction in the form of "conversation". The principal features of this system are it is adaptive and is based on reflection, observation and relation. The system acquires its intelligence through the finite state machine and rule-based agents.