Keyword-based Socially Tenuous Group Queries.
Huaijie Zhu,Wei Liu,Jian Yin,Ningning Cui,Jianliang Xu,Xin Huang,Wang-Chien Lee
DOI: https://doi.org/10.1109/icde55515.2023.00079
2023-01-01
Abstract:Socially tenuous groups (or simply tenuous groups) in a social network/graph refer to subgraphs with few social interactions and weak relationships among members. However, existing studies on tenuous group queries do not consider the user profiles (keywords) of the members whereas in many social network applications, e.g., finding reviewers for paper selection and recommending seed users in social advertising, keywords also need to be considered. Thus, in this paper, we investigate the problem of keywords-based socially tenous group (KTG) queries. A KTG query is to find top N tenuous groups in which the members of each group jointly cover the most number of query keywords. To address the KTG problem, we first propose two exact algorithms, namely KTG-VKC and KTG-VKC-DEG, which give priority to the valid keyword coverage and the combination of valid keyword coverage and degree, respectively, to select members to form a feasible group by adopting a branch and bound (BB) strategy. Moreover, we propose keyword pruning and k-line filtering to accelerate the algorithms. To yield diversified KTG results, we also study the problem of diversified keywords-based socially tenous group (DKTG) queries. To deal with the DKTG problem, we propose a DKTG-Greedy algorithm by exploiting a greedy heuristic in combination with KTG-VKC-DEG. Furthermore, we design two alternative indexes, namely NL and NLRNL, to efficiently check whether the social distance of any two members is greater than the social constraint k in the above algorithms. We conduct extensive experiments using real datasets to validate our ideas and evaluate the proposed algorithms. Experimental results show that the NLRNL index achieves a better performance than the NL index.