Description
Date depot: 1 janvier 1900
Titre: Finding Interesting Events in Twitter and Communities in Social Networks
Directeur de thèse:
Mauro SOZIO (LTCI (EDMH))
Domaine scientifique: Sciences et technologies de l'information et de la communication
Thématique CNRS : Non defini
Resumé:
When events such as the elections to the european parliament or the recent crisis in Ukraine unfold, social network users of Google Plus, Facebook or Twitter engage in intense social activity by establishing new friendship links and posting tweets related to the event. One could represent tweets and social networks as graphs (called entity and social graphs, respectively), where nodes represent entities in a tweet (such as `Obama` or `Crimea`) or users in a social network, while edges measure the strength of the correlation between the corresponding entities in Twitter or denote friendship links in a social network. We can then observe that such events trigger the quick emergence of dense subgraphs in the entity and social graph, that is, subgraphs containing a relatively large number of edges.
In this PhD proposal, we aim at developing efficient algorithms for automatically finding interesting events in Twitter and communities in social networks by finding dense subgraphs in the entity and social graph. This poses several non-trivial challenges, given the sheer size of the graphs described above. Moreover, such graphs are highly dynamic as many friendship links can be established within a relatively short time window and up to 5 thousands tweets per second are posted in Twitter. Therefore, it is crucial to develop algorithms that quickly compute new solutions as the input graph changes over time. In order to cope with the sheer size of the input graph, the PhD candidate will implement these algorithms in modern architectures such as MapReduce, multi-core architectures etc. and evaluate their effectiveness on large real-world graph.
Doctorant.e: Balalau Oana Denisa