Abstract: Social networks play an essential role in the online information diffusion. The understanding of the social relationship among users is a factor beneficial to various applications like viral marketing and advertising through social networks. The advertisers aim to select initially a small number of information hubs and provide free samples of a new product to influence a large number of people to buy the product in a short time. Estimating the influence of users, discussing different domains like daily charts, movies, celebrities and sports require analysis of short tweets posted to a social network. With this intent, the fundamental motive is to maximize the influence of a social network in a certain domain of interest. This work proposes support of self-centered network on domain specific information hub classification (speed-diffusion) technique with the aim of classifying information hubs in different domains. Speed-diffusion is an unsupervised classification model, where the output is based on the software analysis without using training samples. The speed-diffusion technique integrates the self-centered network and N-gram classification model to classify the domain specific information hubs. Speed-diffusion represents each domain as a self-centered network, automatically built using WordNet and Yago Ontology. Analysis of short tweets posted to a social network quantifies the general influence of each user with respect to the self-centered networks of various categories. The N-gram classifier leverages these short tweets posted by the user depending on how far its words are from the self-centered network to mine the information hubs specific to a domain. Finally, it evaluates the accuracy of the speed-diffusion technique than that of the existing approaches using a sample of the Twitter network.