Robust Adaptive Discourse Parsing for E-Learning Fora

Abstract

Discussion fora from e-learning studentspsila platforms were parsed as collective discourse directed either by a specific task achievement goal or by a communication goal. Agora, a robust tool for French and English, provides ldquoforum tilingrdquo but also ldquodiscussion nestingrdquo. It uses the structure of messages along with stylistic statistical data to segment and organize the content of discourse. No external re-sources are needed. The output of this adaptive parser is a scalable view on collective discussion. Such views are used to browse and navigate in a very large forum, or to compare discussion progress between learners groups. Results are presented and their usefulness is discussed.