Welcome

Who is the Proxteam?

Proxteam is an informal research group interested in Computer Science, Linguistics and Cognitive Science. Despite a variety of affiliations and status (see who we are), we unite in a common passion for the study of large terrain networks, especially lexical networks. Most of us work for two CNRS research labs of the University of Toulouse, in France: CLLE, and more specifically its ERSS branch, and IRIT. However, our diverse collaborations strongly connect us to a national and international research network.

Research Interests:

The multidisciplinary research of TerrNet focuses on understanding and leveraging typical features of large terrain networks, especially graphs that originate from linguistic, cognitive or web data.

Large terrain networks model relations between data entities studied in a great variety of fields of interest: epidemiology, economy, sociology, linguistics (semantic networks...), psychology (semantic associations), biology (neural networks, protein interaction networks), information science (hyperlinks between web pages...).

Such networks, despite a great varity of origins, exhibit a remarquably consistent set of properties: they all are Hierarchical Small Worlds networks (HSWs). They have a low edge density but very short average shortest paths, they have edge-dense zones and the degree of their vertices is distributed according to a power-law. This class of HSW networks is however very small relatively to the whole set of possible graphs: the probability, for a network chosen randomly from all possible graphs, of being HSW is very small. That networks that model relations between data entities of our daily lives all share properties otherwise probabilistically rare is quite intriguing.

We seek to understand why and how such structures emerge, what they entail for algorithmics, linguistics and cognitive science and to exploit their specific properties as efficiently as possible in order to approach the complexity of our environment with the most fecund perspective possible.