Distance-based dialog acts labeling

Abstract

The capability to recognize a communication function (dialogue act) of the particular utterance spoken by the speaker in the dialogue interaction, is one of the important ability of human participants in dialogue interactions. Automatic dialog act recognition can make human-computer dialogue interaction more natural and effective. Dialogue act recognition is a language-dependent technology and resources in desired language are required in form of a corpus, where each utterance is labeled by one of the dialog acts. Approaches based on statistical modeling are usually applied to classify dialogue acts. In the paper, simple distance-based approach is introduced and results of evaluation experiments on newly-prepared corpus in Slovak language are presented.