With the rapid development of location-based social networks (LBSNs), multimedia topic modeling on location-related user generated contents (UGCs) for venues is strongly desired. However, most of the previous topic modeling approaches only work on single modality data, or correlated multimodal data. The intrinsic property of UGCs in LBSNs that the heterogeneous UGCs are generally independent makes these approaches unsuitable for multimedia venue topic modeling. In this paper, we propose a novel multimedia topic modeling approach for extracting venue semantics from heterogeneous location-related UGCs. The approach relates multimedia UGCs by leveraging on multiple data sources. Furthermore, a graph clustering method is proposed to detect the topics which are considered as the dense subgraphs. Based on the multimedia venue topic modeling, we further propose the semantic based venue summarization, which verifies the effectiveness of the proposed framework. The integration of these heterogeneous UGCs into semantic topics provides users with an easier way to understand the venues, and therefore enriches the user experience. Extensive experiments have been conducted on a cross-platform dataset and promising results have been obtained. © Springer International Publishing Switzerland 2013.
CITATION STYLE
Nie, W., Wang, X., Zhao, Y. L., Gao, Y., Su, Y., & Chua, T. S. (2013). Venue semantics: Multimedia topic modeling of social media contents. In Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) (Vol. 8294 LNCS, pp. 574–585). Springer Verlag. https://doi.org/10.1007/978-3-319-03731-8_53
Mendeley helps you to discover research relevant for your work.