Towards well-founded extractions for large information diffusion on social networks

Social media platforms have exponentially grown over the past decade. Social networks platforms are used by one-in-three people on the planet, and more than two-thirds of all internet users. This explosive ascending number of users consequently lead to a growth in research of information analysis fields. The purpose of our study is to develop a method that helps propagate information efficiently throughout the social network. To manage this, it is crucial to evaluate user profiles accordingly to their shared content. The principle behind our method is that a user is more likely to share an information, if they are used to sharing content semantically related to its concept. In order to achieve this, this paper focuses on developing an algorithm, which selects seed nodes that are most likely to spread the word across the given social network. The key aspect of our method is to analyze textual content provided within a defined closed social network, while taking into consideration the relevancy aging time of content within the network. For the purpose of this analysis and given the random nature of posts shared by users, we develop our own NED (Name Entity Disambiguation) method, and UWI (Unusual Word Identification) method. We also create a semantic similarity method, in order to match an information with the right users.