This paper presents new conversation modules developedfor TutorJ to support the use of meta-cognitive strategiesin the learners. TutorJ is an Intelligent Tutoring Systemable to interact with users to assess their skills, and improvetheir knowledge in a specific domain. New models are devotedto plan the conversation and to acquire new knowledge fromsemi-structured data sources. Conversation planning is basedon probabilistic techniques. Knowledge acquisition is based onretrieval, transformation and alignment of data from wikis toenrich the replies to students questions.
|Number of pages||10|
|Publication status||Published - 2009|