Local Experts Finding Across Multiple Social Networks

The local experts finding, which aims to identify a set of k people with specialized knowledge around a particular location, has become a hot topic along with the popularity of social networks, such as Twitter, Facebook. Local experts are important for many applications, such as answering local information queries, personalized recommendation. In many real-world applications, complete social information should be collected from multiple social networks, in which people usually participate in and active. However, previous approaches of local experts finding mostly focus on a single social network. In this paper, as far as we know, we are the first to study the local experts finding problem across multiple large social networks. Specifically, we want to identify a set of k people with the highest score, where the score of a person is a combination of local authority and topic knowledge of the person. To efficiently tackle this problem, we propose a novel framework, KTMSNs (knowledge transfer across multiple social networks). KTMSNs consists of two steps. Firstly, given a person over multiple social networks, we calculate the local authority and the topic knowledge, respectively. We propose a social topology-aware inverted index to speed up the calculation of the two values. Secondly, we propose a skyline-based strategy to combine the two values for obtaining the score of a person. Experimental studies on real social network datasets demonstrate the efficiency and effectiveness of our proposed approach.