In face-to-face interaction, the messages exchanged among participants are not limited to chains (patterns) of words, but are characterized by a continuous interaction among information expressed through multiple channels. In addition to the language, in fact, other faculties such as hearing and sight are involved in the construction of meaning: we decode not only the sequences of words, but also the "way" in which they are produced (intonation, volume, speed) and how we move and show ourselves when we pronounce them. The conveyed meaning depends on not only the combination of these characteristics, but also the involvement or the effect that the message has on the recipient(s). Is it possible to predict this effect automatically? In this seminar a new multimodal dataset will be presented for the study and prediction of the potential for the involvement in the context of guided visits to sites of cultural interest.