Longitudinal Detection of Contributor Teams in Wikipedia - Clustering Algorithms in Temporal Graphs

The Lab-STICC, with it affiliation to the INS2I institute of the CNRS, is a research unit historically recognized in Brittany and in France in the field of ICTS. It has a proven capacity to cover a broad scientific spectrum around digital sciences, and in particular with this ability to address various disciplinary fields (Information Theory, Waves & Materials, Embedded Electronics and Computing, Data Sciences, Communication and Signal Detection, Human-Machine Interfaces,...) following multiple themes/application sectors: maritime environment, communicating objects, defense, space, health, security, robotics...

The Lab-STICC's motto "From sensors to knowledge: communicating and deciding" initially underlined the intention to give "meaning" to the "sign", in terms of interpretations and increased value with respect to treatments based on a signal in all its forms. Today, this motto is even more reinforced by the importance of data, which is becoming the vector of technological and scientific innovation, undeniably boosted by the major advances in science in the field of cognition and artificial intelligence. The ability of the Lab-STICC to establish this link between the lower and upper layers (up to Human-Complex Systems cooperation) is an asset to mobilize its scientific strengths, and federate research around the societal challenges such as digital, environment, energy. The laboratory has been restructured around 9 research departments, bringing together 26 teams. It is indeed an architecture based on disciplinary competences - the teams - working in a coordinated way within poles following shared scientific ambitions - and able to articulate in a transversal way around the societal and scientific challenges.

This Master 2 internship aims to develop and evaluate algorithms for longitudinal detection of contributor teams on Wikipedia. The challenge is twofold: on the technical level, it involves designing temporal graph clustering methods capable of tracking the coherent evolution of communities over time; on the application level, these methods must enable analysis of collaborative dynamics on Wikipedia by integrating organizational metrics from management science.

Wikipedia, with over 50 active language versions and millions of contributors, constitutes a privileged observation field for understanding how contributor teams form, evolve, and coordinate within editorial projects. The algorithms will be evaluated on academic benchmarks and then applied to real Wikipedia projects to characterize the formation, evolution, and dissolution of editorial teams.

The internship will focus on implementing and evaluating the Spatio-Temporal Graph Laplacian algorithm, comparing it to baseline methods, and characterizing detected communities using organizational metrics (internal cohesion, collective viability, collaborative effectiveness, core-periphery structure). An essential objective is producing a scientific article for an international conference and releasing all developed code as open source.

This internship is part of the PEPR eNSEMBLE - CONGRATS project (COllectives for kNowledge pRoduction mAnagement That Scale).

Full description of the offer: https://kdrive.infomaniak.com/app/share/396506/f77f1efd-2add-4966-81c3-575827184b43

We are seeking a Master 2 Research student with the following skills:

• Graph algorithmics
• Proficiency in Python (NetworkX, SciPy, pandas)
• Autonomy and organizational capacity