Collecting data for socially intelligent surveillance and monitoring approaches: The case of conflict in competitive conversations

Abstract

One of the most recent trends in surveillance research is the application of socially aware approaches, i.e. approaches that integrate human sciences findings in order to better understand, model and predict the behaviour of people under observation. One of the key requirements for the development of such approaches is the collection of corpora that provide sufficient and reliable information about social phenomena of interest. However, the computing community still pays relatively little attention to the application of methodologies suitable for observational data collection, possibly inspired by human sciences experimental work. This paper tries to fill, at least partially, such a gap by providing an introduction to data collection techniques applied in nonverbal behaviour research. The collection of a corpus aimed at the study of conflict in conversations is used as a case study and example.

Holistic Affect Recognition Using PaNDA: Paralinguistic Non-Metric Dimensional Analysis

“…The SSPNet Conflict Corpus (SCC) [63] was extracted from the Canal 9 Corpus [74], a collection of Swiss political debates. Each clip was rated in terms of the intensity of conflict on an interval scale [−10, +10]via Amazon Mechanical Turk, using a questionnaire including physical (objective observation) and inferential (subjective interpretation) questions [75].…”

Random Discriminative Projection Based Feature Selection with Application to Conflict Recognition

“…It contains political debates televised in Switzerland 1 . The statistics of the corpus are summarized in Table II. The clips have been annotated following the process illustrated in [29] with respect to conflict level by roughly 550 assessors recruited via Amazon Mechanical Turk. Each clip is assigned a continuous conflict score in the range , giving rise to a regression task.…”

Predicting Continuous Conflict Perceptionwith Bayesian Gaussian Processes

“…The literature proposes a large number of approaches dealing with some of the most important social and psychological phenomena (see [6] for an extensive survey), but conflict has received only limited attention because it is difficult to access ecologically valid data [26]. For this reason, earlier works focused on agreement and disagreement, easier to observe and annotate, while actual conflict detection and measurement approaches appeared only recently.…”