Analyse des réseaux sociaux | EPFL Graph Search

L'analyse des réseaux sociaux est une approche issue de la sociologie, qui a recours à la théorie des réseaux afin d'étudier les interactions sociales, en termes de réseau. La théorie des réseaux sociaux conçoit les interactions sociales en termes de nœuds et liens. Les nœuds sont habituellement les acteurs sociaux dans le réseau, mais ils peuvent aussi représenter des institutions, et les liens sont les interactions ou les relations entre ces nœuds. L'analyse des réseaux sociaux permet ainsi de prendre connaissance des acteurs clés du réseau, de l'influence de chacun dans le groupe et de la qualité des interactions et des relations qu'ils entretiennent. Il peut exister plusieurs sortes de liens entre les nœuds. Dans sa forme la plus simple, un réseau social se modélise pour former une structure analysable où tous les liens significatifs entre les nœuds sont étudiés. Il en va de même pour les trous structuraux, c'est-à-dire une absence de liens directs entre deux sommets. Il est ent

Three Essays on Methodologies for Dynamic Modeling of Emerging Socio-technical Systems

Mohammad Amin Dehdarian

Socio-technical energy transitions are long-term and major transformations in incumbent energy infrastructures. They include fundamental changes in technologies as well as institutions and social patterns. Transition studies are primarily focused on frameworks for analyzing the entire transition process by investigating the historical cases of transitions. A multi-phase approach to transition posits this process begins with a pre-development phase characterized by technological and institutional lock-ins, and resistance from incumbent actors. This period is critical for a forward-looking approach to transitions, since early developments shape path-dependent and irreversible processes leading to the emergence of new transition pathways. However, our understanding about the mechanisms and dynamics of this phase is still very limited. This is mainly due to lack of data, weak conceptualization and the necessity of developing new methods proper to deal with these limitations. This dissertation develops methodologies for investigating some complex questions arising in the pre-development phase, by focusing on the case of smart grid development. The first essay uses insights from modeling interventions in complex systems and builds a System Dynamics model to investigate the cost allocation problem of smart metering roll-out. The second essay takes ideas from Technological Innovation System approach and develops a method to analyze the emergence of spatial diversity in smart grid development by combining Social Network Analysis and Agent-Based Modeling. The third essay builds on ideas from network theory and evolutionary modeling to develop a method for identifying the main path of knowledge development and analyzing knowledge trajectories in smart grid initiatives.

EPFL2017

Social Network Analysis in Multimedia Indexing: Making Sense of People in Multiparty Recordings

Sarah Favre

This paper presents an automatic approach to analyze the human interactions appearing in multiparty data, aiming at understanding the data content and at extracting social informa- tion such as Which role do people play?, What is their attitude?, or Can people be split into meaningful groups?. To extract such information, we use a set of mathematical techniques, namely Social Networks Analysis (SNA), developed by sociologists to analyze social interac- tions. This paper shows that a strong connection can be established between the content of broadcast data and the social interactions of the individuals involved in the recordings. Experiments aiming at assigning each individual to a social group corresponding to a specific topic in broadcast news, and experiments aiming at recognizing the role played by each indi- vidual in multiparty data are presented in this paper. The results achieved are satisfactory, which suggests on one side that the application of SNA to similar problems could lead to useful contributions in the domain of multimedia content analysis, and on the other side, that the presented analysis of social interactions could be a significant breakthrough for affective computing.

2009

Social Network Analysis for Automatic Role Recognition

Sarah Favre

The computing community has shown a significant interest for the analysis of social interactions in the last decade. Different aspects of social interactions have been studied such as dominance, emotions, conflicts, etc. However, the recognition of roles has been neglected whereas these are a key aspect of social interactions. In fact, sociologists have shown not only that people play roles each time they interact, but also that roles shape behavior and expectations of interacting participants. The aim of this thesis is to fill this gap by investigating the problem of automatic role recognition in a wide range of interaction settings, including production environments, e.g. news and talk-shows, and spontaneous exchanges, e.g. meetings. The proposed role recognition approach includes two main steps. The first step aims at representing the individuals involved in an interaction with feature vectors accounting for their relationships with others. This step includes three main stages, namely segmentation of audio into turns (i.e. time intervals during which only one person talks), conversion of the sequence of turns into a social network, and use of the social network as a tool to extract features for each person. The second step uses machine learning methods to map the feature vectors into roles. The experiments have been carried out over roughly 90 hours of material. This is not only one of the largest databases ever used in literature on role recognition, but also the only one, to the best of our knowledge, including different interaction settings. In the experiments, the accuracy of the percentage of data correctly labeled in terms of roles is roughly 80% in production environments and 70% in spontaneous exchanges (lexical features have been added in the latter case). The importance of roles has been assessed in an application scenario as well. In particular, the thesis shows that roles help to segment talk-shows into stories, i.e. time intervals during which a single topic is discussed, with satisfactory performance. The main contributions of this thesis are as follows: To the best of our knowledge, this is the first work where social network analysis is applied to automatic analysis of conversation recordings. This thesis provides the first quantitative measure of how much roles constrain conversations, and a large corpus of recordings annotated in terms of roles. The results of this work have been published in one journal paper, and in five conference articles.

EPFL2011

Social Network Analysis for Automatic Role Recognition

Sarah Favre

EPFL2011