Jason Vallet

Labelled Graph Strategic Rewriting for Social Networks

We develop an algebraic approach, based on labelled-graph strategic rewriting , for the study of social networks, specifically network generation and propagation mechanisms. This approach sheds a new light on these problems, and leads to new or improved generation and propagation algorithms. We argue that relevant concepts are provided by three ingredients: labelled graphs to represent networks of data or users, rewrite rules to describe concurrent local transformations, and strategies to express control. We show how these techniques can be used to generate random networks that are suitable for social network analysis, simulate different propagation mechanisms, and analyse and compare propagation models by extracting common rules and differences, thus leading to improved algorithms. We illustrate with examples the flexibility of the approach.

Labelled Graph Rewriting Meets Social Networks

The intense development of computing techniques and the increasing volumes of produced data raise many modelling and analysis challenges. There is a need to represent and analyse information that is: complex –due to the presence of massive and highly heterogeneous data–, dynamic –due to interactions, time, external and internal evolutions–, connected and distributed in networks. We argue in this work that relevant concepts to address these challenges are provided by three ingredients: labelled graphs to represent networks of data or objects; rewrite rules to deal with concurrent local transformations; strategies to express control versus autonomy and to focus on points of interests. To illustrate the use of these concepts, we choose to focus our interest on social networks analysis, and more precisely in this paper on random network generation. Labelled graph strategic rewriting provides a formalism in which different models can be generated and compared. Conversely, the study of social networks, with their size and complexity, stimulates the search for structure and efficiency in graph rewriting. It also motivated the design of new or more general kinds of graphs, rules and strategies (for instance, to define positions in graphs), which are illustrated here. This opens the way to further theoretical and practical questions for the rewriting community.

Semantic Social Networks: A New Approach to Scaling Digital Ethnography

We propose a data-based approach to doing ethnographic research in a digital environment. It has three main components. First, it treats online conversational environments as human communities, that ethnographers can engage with as they would in onsite fieldwork. Second, it represents those conversations, and the fieldnotes made by researchers thereon, in network form. We call these networks semantic social networks, as they incorporate information on social interaction and their meaning. They encode a map of the associations between key concepts as perceived by informants as a group. Third, it uses methods borrowed from network science to process these data.

JASPER: Just A new Space-filling and Pixel-oriented layout for large graph ovERview.

When analysing data and handling a visualisation, users mainly spend their cognitive resources making sense of the graph-ical representation and mapping it back to the data and domain. This task becomes even more critical when dealing with larger data sets. Therefore, a valuable visualisation design strategy is to couple graphical representations and user tasks to better support the sense making process. This paper focuses on a particular task where users must make sense of state changes occurring on nodes of a graph. To this end, we propose JASPER, a new layout algorithm focusing on the visualisation of nodes inspired from pixel-oriented layouts, relying on node clustering to identify and represent existing connections through spatial adjacency. JASPER can layout moderate size graphs in real-time and is able to tackle large graphs with up to 2 million nodes and 5 million edges in reasonable time (about half a minute). Furthermore, although JASPER has been designed around a specific application , the underlying methodology can be employed to draw quick overviews of any type of graphs. The paper lays down the underlying principles of JASPER, and reports it performances (execution times) on increasingly large graphs. JASPER is then used and showcased to visualise network propagation phenomenon in large graphs.

Studying propagation dynamics in networks through rule-based modeling

Modeling propagation dynamics on networks is an amazingly fertile and active area of research. Roughly speaking, network models aim at gaining a better understanding of how actors influence the overall network behaviour through their individual actions. However, considering the extended literature surrounding the subject, one is entitled to think that moving beyond the state-of-the-art in network modeling requires the ability to compare models, or consider slight variations of a model. This requires having a common language describing all considered models, allowing to objectively compare them and unfold their inherent properties and complexity. This also assumes users can easily run models, steer them and interactively evaluate their performance and behaviour. The approach we describe aims at providing a framework turning network propagation modeling into rule-based modeling (aka graph rewriting). That is, models are described as a set of algorithmic transformation rules acting locally. Our approach has partially been validated by providing such a description of a well-known model relying on probabilistic rules, where nodes trigger actions depending on their neighbors influences. The results so obtained confirm rule-based modeling as a promising avenue. The use of a visual analytics framework to conduct such tasks is vital and motivated us to further develop and adapt a general purpose visual analytics system for graph rewriting to the particular case of network propagation.