"Love ya, jerkface": Using Sparse Log-Linear Models to Build Positive and Impolite Relationships with Teens

摘要: One challenge of implementing spoken dialogue systems for long-term interaction is how to adapt the as user and system become more familiar. We believe this includes evoking signaling aspects relationships such rapport. For tutoring systems, may additionally require knowing are signaled among non-adult users. therefore investigate conversational strategies used by teenagers in peer dialogues, these function differently friends or strangers. In particular, we use annotated automatically extracted linguistic devices predict impoliteness positivity next turn. To take into account sparse nature features real data models including Lasso, ridge estimator, elastic net. evaluate predictive power our under various settings, compare with standard non-sparse solutions. Our experiments demonstrate that accurate than quantitatively, teens unexpected kinds language do relationship work rapport, but strangers, tutors tutees, carry out quite different ways from one another.