The use of text mining to build a pedagogical agent capable of mediating synchronous online discussions in the context of foreign language learning

The present paper discusses the use of text mining to support the design of a pedagogical agent that mediates synchronous online discussions of academic texts by undergraduate students of English as a foreign language. The pedagogical agent proposed here has the instructional role of a tutor collaborator that participates in the chat discussion following mediation strategies grounded on sociocultural theory to assist the collective writing. Furthermore, we propose a pedagogical agent model that uses text mining techniques to identify when students deviate from the main topics that should be addressed in their discussions in a real time chat. Another important function enabled by the use of the text mining tool involves the assessment of the discussion relevance in relation to the base text, which supports the pedagogical agent decision towards a more adequate intervention. The conception of using text mining to guide the pedagogical agent in the mediation of the students´ discussions has been based on previous research that has already shown how this particular mining tool could support educators´ work in the evaluation of essays and of students´ contributions in discussion forums. Preliminary results of this study are also presented showing the agent´s potential to foster students´ online conversations.