| dc.description.abstract | Identifying whether an utterance is a statement, question, greeting, and so forth is integral to effective automatic understanding of natural dialog. Little is known, however, about how such dialog acts (DAs) can be automatically classified in truly natural conversation. This study asks whether current approaches, which use mainly word information, could be improved by adding prosodic information. The study examines over 1000 conversations from the Switchboard corpus. DAs were handannotated, and prosodic features (duration, pause, F0, energy and speakingrate features) were automatically extracted for each DA. In training, decision trees based on these features were inferred; trees were then applied to unseen test data to evaluate performance. For an allway classification as well as three subtasks, prosody allowed highly significant classification
over chance. Featurespecific analyses further revealed that although canonical features (such as F0 for questions) were important, less obvious features could compensate if canonical features were removed. Finally, in each task, integrating the prosodic model with a DAspecific
statistical language model improved performance over that of the language model alone. Results suggest that DAs are redundantly marked
in natural conversation, and that a variety of automatically extractable prosodic features could aid dialog processing in speech applications. | en |
