Machines that detect and classify human emotions should have a better read on a user’s mental state, while machines that display human emotions should be able to express more subtle machine states. A competent, bidirectional emotional interaction should therefore communicate with users more effectively. But how? This talk discusses Balentine’s research into emotional speech behavior and its effect on artificial conversation. The goal is to identify specific user-interface design methods that make positive use of emotion.

Speech can reveal the speaker’s emotional condition, sentiment, or attitude on a specific topic and the impact of a public speaker on an audience. We review different ways to measure the human mental state and how it is projected through speech. We cover both text-based analysis and signal-based evaluation. We explain the differences between emotion detection, sentiment analysis, and persona classification as well as the technology behind each view into a speaker’s psyche and how this information can deliver business returns.