Virtual sighs to understand human sighs

Various vocalizations are displayed in everyday conversation and TV debates and talk shows, with or without communicative import: cough, hiccup, laughter. While laughter has been the object of intense research, a peculiar vocalization has been less investigated: the sigh. While Boncinelli (2012) and other authors explained its structure as a peculiar kind of breath, only two studies have analyzed its semantic and pragmatic functions (Teigen, 2008; Hoey, 2015). The aim or this work is to outline a typology of sighs and to distinguish its various meanings while connecting them to slight differences in the ways they are physically produced. To do so, we exploited various methods. First we collected a corpus of sighs from yoututbe videos drawn from movies, cartoons, fiction, talk shows and political debates. Then we classified each sign in terms of its signal whether including inspiration, expiration, or both, and whether audible or non-audible), and in terms of its meaning boredom, insufference, weariness, sorrow, annoyance, frustration, longing, relief...). Finally we simulated different types of sigh in the Virtual Agent Maya (Origlia and Cosi, 2016) and in a perception study we investigated if subjects find out a correspondence between specific patterns of the signal and specific meanings.