Who are there: Discover Twitter users and tweets for target area using mention relationship strength and local tweet ratio. (March 2022)

• Record Type: Journal Article
• Authors: Liu, Yimin; Luo, Xiangyang; Zhang, Meng; Tao, Zhiyuan; Liu, Fenlin
• Is Part Of: Journal of network and computer applications. Volume 199(2022)
• Publication Date: 2022-03
• Subjects: 00-01 -- 99-00; Local tweet ratio -- Mention relationship strength -- Twitter -- User discovery; Microcomputers -- Periodicals; Computer networks -- Periodicals; Application software -- Periodicals; Micro-ordinateurs -- Périodiques; Réseaux d'ordinateurs -- Périodiques; Logiciels d'application -- Périodiques; Application software; Computer networks; Microcomputers; Periodicals; 004.05; 004
• DOI: 10.1016/j.jnca.2021.103302
• Languages: English
• ISSNs: 1084-8045

Abstract

Abstract: The goal of Twitter users and tweets discovery for the target area is to find resident users and tweets created in the area. Previous works usually utilize following relationships among users to discover new users, whose precision and speed are easily affected by fake followers and the rate-limiting of getting Twitter data. In this manuscript, an algorithm for users and tweets discovery based on mention relationship strength and local tweet ratio (MRS-LTR for short) is proposed. First, the initial tweets are obtained by using Twitter API, from which the seed users and the seed tweets are extracted, and the mention relationship graph is constructed. Then, two sound-reasonable hypotheses about the location information contained in mention relationship and user's tweets are explored and verified on the public dataset. Next, based on the two hypotheses, three new recommendation indicators are proposed to recommend new users in the area from the users mentioned by seed users. Finally, the candidate user location verification method based on profile and iteration probability is used to expand the seed user set. The principle analysis of discovery precision, discovery speed and algorithm function shows the effectiveness of MRS-LTR. Experimental results on a dataset of 4.4M users and 112M tweets demonstrate that MRS-LTR outperforms the state-of-the-art algorithm. The precision infimum and modified precision are increased by 14.79% and 31.22%, respectively. Meanwhile, the discovery speed is significantly improved, and the number of users discovered in the same period is about 8.7 times that of the existing algorithm.