Linking speaking and looking behavior patterns with group composition, perception, and performance

Abstract

This paper addresses the task of mining typical behavioral patterns from small group face-to-face interactions and linking them to social-psychological group variables. Towards this goal, we define group speaking and looking cues by aggregating automatically extracted cues at the individual and dyadic levels. Then, we define a bag of nonverbal patterns (Bag-of-NVPs) to discretize the group cues. The topics learnt using the Latent Dirichlet Allocation (LDA) topic model are then interpreted by studying the correlations with group variables such as group composition, group interpersonal perception, and group performance. Our results show that both group behavior cues and topics have significant correlations with (and predictive information for) all the above variables. For our study, we use interactions with unacquainted members i.e. newly formed groups.

Personality Trait Classification via Co-Occurrent Multiparty Multimodal Event Discovery

“…Head gesture and gaze are approximated using head posture and motion. [10] points to visual focus of attention as a key feature to link impressions from group members. [17] also points to gaze states as effective feature to detects important statements.…”

“…On the other hand, unsupervised learning approaches can often find intermediate representations that help understand these higher concepts [10].…”

Predicting Performance of Collaborative Storytelling Using Multimodal Analysis

“…That research focuses on a listener's head gesture recognition in a dyadic interaction. [21], [22] use a latent Dirichlet allocation (LDA) model for mining context features in group. In [22], group features for speaking status and gaze state are used for input to an LDA.…”

“…Conversely, we used the various kind of feature set not only speech and gaze cues, but also hand and head gesture features for modeling the performance. Main objective of [21], [22] is finding context features to link role of participant or personality trait. Our objective is modeling to predict storyteliing performance using context features and is quite different with these objective.…”

“…However, although the feature set exhibits significant linear correlation with the performance, it is not always the most effective feature set to predict it; the correlation value can be used as a criterion for feature selection [27] in machine learning. This analysis is also important in order to understand the relationship between each set of variables, and it was widely used in many researches for this purpose [12], [22] to understand it. These results of the pairwise correlation are also important as a reference for a future work to develop a model to predict other kinds of performance (explanation in different setting).…”