CYBEREMOTIONS WORKSHOP Emergent Emotions on the WebJozef Stefan Institute, Ljubljana: 20-21. January 2011A B S T R A C T S |
| SPEAKER | ADDRESS | T i t l e & A b s t r a c t |
| Bosiljka Tadic | Department of Theoretical Physics, Jozef Stefan Institute, Ljubljana | Opening remarks & Project activity from the Networks' view I put some remarks regarding collaboration of WP6 with others, with the emphasis on providing data, modeling, or using theoretical models as an input.
|
| Julian Sienkiewicz | Faculty of Physics, Warsaw University of Technology Koszykowa 75 Warsaw 00-662 Poland | WP1 report: Model of emotional dialogues based on entropy growth We propose a way to model bot-human interactions based on the statistical analysis of human-human data from Ubuntu discussion groups. The common discussion is decomposed into several one-to-one dialogues, which possess similar properties. The average number of positive and objective comments equalize close to the end of discussion and the difference between initial and terminal value of entropy is at the same level for all discussion lengths. Based on those facts as well as on the observed cluster growth we construct a model producing a distribution of discussion lengths similar to the one observed in the data. We also give hints for the conditions that result in sustaining dialogue duration. |
| Stephane Gobron | EPFL IC ISIM VRlab, Station 14 Lausanne CH-1015 Switzerland | WP2 report: 3D emoChattings Based on our previous word-to-graphical-emotion work, we have elaborated two models and respective 3D software to simulate verbal and non-verbal communication between virtual humans. The resulting interpersonal micro virtual societies were tested by two user-tests: first at individual level with avatar-to-agent relationships; second, ¬at social behavior level using multiple avatar relationships with 8 simultaneous conversations in a single virtual world. |
| Georgios Paltoglou | University of Wolverhampton Wulfruna Street, Wolverhampton, WV1 1LY, UK | WP3 report: Progress within WP3In this talk we will discuss the progress on sentiment analysis, data gathering and data analysis that has taken place within the group within the last months. COAUTHORS: (Lexicon/ANEW classifier) George,
(Twitter analysis) Mike and (Newgroup dataset) Hannes. |
| Marcin Skowron | OFAI Freyung 6/3/1a Vienna 1010 Austria | WP4 report: Effect of an Affective Profile in Dialog System-User Conversations We present findings from our recent experiments on the effect of conversant affective profile and its influence on the communication processes. In particular, we report on Internet users subjective evaluation of artificial conversational entities that varied in their affective profile i.e., positive, negative, neutral. |
| David Garcia | ETH Zurich Kreuzplatz 5 Zurich 8032 Switzerland | WP5 report: Product Reviews and IRC Channel Analysis The first applications of the theoretical framework for collective emotions
developed by ETHZ are summarized. Using an improved amazon.com dataset with
valence and news information, we design a new model to explain some features of
consumer online emotions. In parallel, we analyzed the IRC chatroom data. We
study the distribution of waiting times for user and channel utterances, to find
evidence of a known model from queue theory that tells us about the timing of
real-time Internet interaction. Using Detrended Fluctuation Analysis, we have
found the distribution of individual emotional persistence, which shows the
heterogeneous ways in which people keep or change the emotions they express.
This results can serve as directives for emotional dialog systems design.
This is a joint work with COAUTHORS: Antonios Garas and Frank Schweitzer. |
| Marija MItrovic | Department of Theoretical Physics, Jozef Stefan Institute, Ljubljana | WP6 report: Modeling of Emotional Agents on Blogs Modeling of the emotional interactions between humans in on-line world is an
important step towards understanding of the dynamics on Blogs and similar sites. We have developed a model of strictly emotional agents, who are interacting on growing weighted bipartite networks of posts and users (agents). The emotional state defined via two local variables, arousal and valence of an agent, is fluctuating in time, depending on the emotional contents of posts through which the agents are connected and several control parameters. Some control parameters can be derived from the empirical data from Blogs, Diggs etc.
Our results show that the user communities are formed as a consequence of their
emotional interactions over time. Detailed analysis of the emotional contents of each community and other measures of the collective behavior of users are done by simulations for different parameters. COAUTHORS: B. Tadic, M. Suvakov, G. Paltoglou, D. Garcia
|
| Schahram Dustdar | Distributed Systems Group, Information Systems Institute, Vienna University of Technology | (INVITED LECTURE & Colloquium IJS) Self-Adaptation in Socio-Computational Future Internet Systems. In this presentation i will present some of our work in building self-adaptive systems and in particular discuss some techniques and mechanisms we employ to achieve socially enhanced adaptation in service-oriented systems.
|
| Elena Tsankova | Jacobs University Campus Ring 1 Bremen 28759 Germany | WP7 report: Emotional reactions while reading and writing on Internet forums: A WP7 data set We will provide an overview of the prospective contents of our second data deliverable (D7.2), including the experimental design of the data collection as well as some initial results. In this second stage of the project, we assessed human emotional responses to active participation in online forums. For this purpose we conducted a study where participants generated text in an environment resembling an Internet forum, while repeated measures of their subjective emotional state were taken and three indicators of their psychophysiological responses (facial electromyography, bilateral electrodermal activity, and electrocardiography) were recorded continuously. Data was collected from a total of 65 participants, and for each measure, data of at least 60 participants could be retained after removal of artefacts and technical malfunction. The data collected in this study builds upon the results of our previous studies on perception of Internet forum posts and allows comparisons between subjective and physiological measures during different types of events, for example, writing a beginning of a forum thread, writing a response to an already existing thread, thinking about the topic one is going to write about, writing and thinking about positive and negative topics. As in our previous deliverable (D7.1), data on responses to IAPS images has again been included for comparison purposes. COAUTHOR: M. Theunis |
| Robert Hillmann | TU Berlin Franklinstrasse 28-29 Berlin 10587 Germany | WP8 report: Dynamic Network Visualization and Network Sentiment Analysis In addition to developing a dynamic network visualization viewer, WP 8 continued to convert further datasets into event-based networks. Examples include IRC-Chat data and Newsgroup data provided by our project collaborators. The IRC-Data was sliced into monthly subnetworks to maintain a reasonable network size. As a result, the converted datasets contain 1536 social networks from the IRC-Data and 1693 networks from the Newsgroup data.
In line with our work plan and based on the above steps, WP8 then advanced the work on link based network sentiment analysis with statistical methods. For that, we developed and adapted a triad census analysis approach for our longitudinal event-based network data and performed a series of analyses of triadic sentiment dissemination patterns across different data sources.
Additionally, we have started to complement this link based analysis with an actor based approach, where we investigate the impact of sentiment-laden communication events on the emotional valence of actor activity. |
| Stefan Thurner | Complex Systems Research Group, Medical University of Vienna | (INVITED LECTURE)Deriving Human Behavioral Laws from a Large-Scale Computer Game Society
The capacity to collect fingerprints of individuals in online media has revolutionized the way researchers explore social systems. Such systems can often be described in terms of large, complex networks that have a topology of interconnected nodes. Much emphasis has been put on the existence or absence of interactions between actors, as well as their possible strength or directionality, but the nature of these interactions has been overlooked in most empirical studies, mainly because it is usually unavailable in large-scale data. In this work, we perform an analysis of a large-scale social network of a computer game society where different types of one-to-one interactions can be identified: friendship, enmity, communication, trade, bounty and war. Based on these networks, we present a series of findings in regularities in human interaction patterns,and the role of interacting social networks in the organization of the social system. We present a first large-scale verification of structural balance theory and provide first results of the nature of streams of human actions - a human behavioral code.
|
| Anna Borowiec | Gemius Woloska 7 Str. MARS Building, Entrance D Warsaw 02-675 Poland | WP9 report: The panelists participation in the experiment EmoDialog 1In December 2010, in cooperation OFAI, WUT and Gemius conducted an experiment EmoDialog1. The role of Gemius was recruiting participants. The experiment lasted 10 days and during that time 3,000 people were invited to participate in the experiment. This presentation will present the characteristics of users who participated in the experiment EmoDialog1. |
| Janusz Debski | Warsaw University of Technology | Project Management, Summary & To Do ...will be added at the end of meeting. IN COLLABORATION with Janusz Holyst, Project coordinator |
| top... | home... | |