Mapping Speech Intonations to the VAD Model of Emotions

Human speech carries information not only in its semantics but also in intonation, that delivers a vast amount of knowledge that is usually ignored in modern text-to-speech (TTS) technologies. Development of methods to define and further to compute dependencies between speech and emotions is considered a promising field of research. After looking into the role of intonation in the Russian language an experiment has been suggested to create a mapping between intonational configurations of spoken Russian language and sentiments. The impact of emotions on individuals is tangible on many levels, that is the reason why they are a key part of modelling affective states. As understanding sentiments is as important as expressing them, ways to define them vary from discrete sets to multidimensional spaces. Therefore the suggested mapping is expected to be used further in the task of synthesising more natural and emotion-saturated speech.