Towards human-machine dialog in Slovak

Abstract

Spoken dialog represents a comfortable way of the human-machine cooperation. Dialogue technology is language-dependent due to its relationship to the natural language processing. For a long time, there did not exist any resources for designing advanced dialogue interaction in the Slovak language. Therefore, the new corpus of human-human dialogues started to be prepared. Suitable dialogue interactions from several TV shows were identified and processed. The corpus contains annotations of dialogue acts (DA) and transcriptions of the spoken content. Dialogue act classification plays an important role in advanced dialogue management systems. They represent the intention of the dialogue participant in the particular part of the dialogue. New simplified annotation schema was designed and used for labeling the corpus with 13 DA labels. Bigram models of DA classes and dialogue grammar were trained on the corpus. The HMM-based approach was applied to perform DA classification of utterances in the Slovak language. The paper also summarizes the state-of-the-art in the area of the human-machine spoken dialog in the Slovak language.