Automatic role recognition based on conversational and prosodic behaviour

Salamin, Hugues; Vinciarelli, Alessandro; Truong, Khiet P.; Mohammadi, Gelareh

doi:10.1145/1873951.1874094

Cited by 9 publications

(5 citation statements)

References 12 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Using the same data and targeting the same roles as above, Salamin et al [23] exploited specific acoustic features, such as who talks when and how much (turn-taking), and statistical properties of pitch, formant, energy, and speaking rate, reporting an accuracy of 89%.…”

Section: B Social Computing Approachesmentioning

confidence: 99%

Modeling Functional Roles Dynamics in Small Group Interactions

Dong

Lepri

Pianesi

et al. 2013

IEEE Trans. Multimedia

View full text Add to dashboard Cite

The paper addresses the automatic recognition of social and task-oriented functional roles in small-group meetings, focusing on several properties: a) the importance of non-linguistic behaviors, b) the relative time-consistency of the social roles played by a given person during the course of a meeting, and c) the interplays and mutual constraints among the roles enacted by the different participants in a social encounter. In particular, this paper proposes that the Influence Model framework can address these properties of functional roles, and compares the performance obtained by this framework to the performances of models that consider only property (a) (SVM), and to those that address both (a) and (b) (HMM). The results obtained confirm our expectations: the classification of social functional roles improves if models account for temporal dependencies among the roles played by the same subject, for the time properties of the roles played by each individual, and for the mutual constraints among the roles of different group members. The two versions of the Influence Model (IM and newIM), which encode all three properties together, outperform both the SVM and the HMM on most of the figures of merit used. Of particular interest is the capability of the Influence Model to obtain good or very good results on the less-populated classes -Orienteer and Seeker for the task area, and Attacker and Supporter for the socio-emotional area.

show abstract

Section: B Social Computing Approachesmentioning

confidence: 99%

Modeling Functional Roles Dynamics in Small Group Interactions

Dong

Lepri

Pianesi

et al. 2013

IEEE Trans. Multimedia

View full text Add to dashboard Cite

show abstract

“…We also have compared our results to prior similar studies on social dynamics of small groups using different technologies. These prior studies commonly collected much richer data (e.g., speaking turn and prosodic cues [100,101], head and body activity [101]) in both visuals and audios, while our tags collect only body distances and orientation. The most relevant work is [117], where Zancanaro et al used cameras and microphones to analyze the roles played by team members in relation to the tasks the group has to face ("Task Area") and in relation to the functioning of the group ("Socio-Emotional Area").…”

Section: Resultsmentioning

confidence: 99%

“…In [54], group cohesion is studied using hours of audio-visual group meeting data. [100] uses prosodic and turn-taking behaviors to identify participant's speaking role. [55] estimates group formations in crowded environments using a graph clustering algorithm.…”

Section: Related Workmentioning

confidence: 99%

Measuring Interaction Proxemics with Wearable Light Tags

Montanari

Zhao

Francu

et al. 2018

Proc. ACM Interact. Mob. Wearable Ubiquitous Technol.

View full text Add to dashboard Cite

The proxemics of social interactions (e.g., body distance, relative orientation) influences many aspects of our everyday life: from patients' reactions to interaction with physicians, successes in job interviews, to effective teamwork. Traditionally, interaction proxemics has been studied via questionnaires and participant observations, imposing high burden on users, low scalability and precision, and often biases.In this paper we present Protractor, a novel wearable technology for measuring interaction proxemics as part of non-verbal behavior cues with fine granularity. Protractor employs near-infrared light to monitor both the distance and relative body orientation of interacting users. We leverage the characteristics of near-infrared light (i.e., line-of-sight propagation) to accurately and reliably identify interactions; a pair of collocated photodiodes aid the inference of relative interaction angle and distance. We achieve robustness against temporary blockage of the light channel (e.g., by the user's hand or clothes) by designing sensor fusion algorithms that exploit inertial sensors to obviate the absence of light tracking results.We fabricated Protractor tags and conducted real-world experiments. Results show its accuracy in tracking body distances and relative angles. The framework achieves less than 6 • error 95% of the time for measuring relative body orientation and 2.3-cm -4.9-cm mean error in estimating interaction distance. We deployed Protractor tags to track user's non-verbal behaviors when conducting collaborative group tasks. Results with 64 participants show that distance and angle data from Protractor tags can help assess individual's task role with 84.9% accuracy, and identify task timeline with 93.2% accuracy.

show abstract

“…The method is based on minimal set of objective speech tags that is beyond annotators' agreement (speakers, acoustic silences, and overlaps), and with no time consuming transcriptions. Methods as we suggest can further contribute to automatic role recognition [28], [29], and [30], to clienttherapist automatic assessment tools, such as The Motivational Interviewing Skills Code (MISC) [31], and to humancomputer interface [32].…”

Section: Resultsmentioning

confidence: 99%

In search of the role's footprints in client-therapist dialogues

Lerner¹,

Silber-Varod²,

Batista³

et al. 2016

Speech Prosody 2016

View full text Add to dashboard Cite

The goal of this research is to identify speaker's role via machine learning of broad acoustic parameters, in order to understand how an occupation, or a role, affects voice characteristics. The examined corpus consists of recordings taken under the same psychological paradigm (Process Work). Four interns were involved in four genuine client-therapist treatment sessions, where each individual had to train her therapeutic skills on her colleague that, in her turn, participated as a client. This uniform setting provided a unique opportunity to examine how role affects speaker's prosody. By a collection of machine learning algorithms, we tested automatic classification of the role across sessions. Results based on the acoustic properties show high classification rates, suggesting that there are discriminative acoustic features of speaker's role, as either a therapist or a client.

show abstract

Automatic role recognition based on conversational and prosodic behaviour

Cited by 9 publications

References 12 publications

Modeling Functional Roles Dynamics in Small Group Interactions

Modeling Functional Roles Dynamics in Small Group Interactions

Measuring Interaction Proxemics with Wearable Light Tags

In search of the role's footprints in client-therapist dialogues

Contact Info

Product

Resources

About