Human/Human Conversation Understanding
Gökhan Tür, Dilek Zeynep Hakkani-Tür · 2011
This chapter focuses on two-party and multiparty human/human conversation understanding, mainly focuses on discourse modeling, speech act modeling, and argument diagramming. It tries to bridge the gap between studies in human/human conversations and human/machine systems in order to either build better human/machine conversational systems or to better understand human/human conversations. While speech is the most natural medium of human/human communication, little spoken data is available for research purposes due to privacy or copyright issues or issues relating to signal quality such as non-ideal recording conditions. The chapter provides an overview of some of the research done in the last decade for processing and accessing the content of human/human conversational speech. Furthermore, as dialogue acts are designed to be task-independent, it is easier to reuse them or use them as a starting point when processing new genres. Controlled Vocabulary Terms human computer interaction; speech processing