A Preliminary Investigation of Learner Characteristics for Unsupervised Dialogue Act Classification.
Aysu Ezen-Can, Kristy Elizabeth Boyer · 2014
For tutorial dialogue systems, classifying the dialogue act (such as questions, requests for feedback, or statements) of student natural language utterances is a central challenge. Recently, momentum is building for the use of unsupervised machine learning approaches to address this problem because they reduce the manual tagging required to build dialogue act models from corpora. However, unsupervised models still do not perform as well as supervised models in terms of accuracy. This paper presents an unsupervised dialogue act modeling approach that leverages the influence of learner characteristics, particularly students ’ perceptions of their own skill, on their language use. The experimental findings show that leveraging skill perception within dialogue act classification improves performance of the models, producing better accuracy. This line of investigation will inform the design of next-generation tutorial dialogue systems, which leverage machine-learned models to adapt to their users.