Web video topics discovery and structuralization with social network

MLLDA: Multi-level LDA for modelling users on content curation social networks

“…Modelling users is also researched for topic modelling of documents and recommendation [18,19,20,21,26]. In [18] a generative author-topic model is proposed for modelling authors for a collection of documents for exploratory analysis of documents.…”

“…In [21] topic analysis of documents and community discovery of authors are brought together in one unified hierarchical Bayesian model. The work in [26] is aimed to discover topics for web videos, considering the social networks formed among users. Collaborative topic modelling for recommending scientific articles is proposed to recommend scientific article to online communities [27].…”

Tracking topic evolution via salient keyword matching with consideration of semantic broadness for Web video discovery

“…Although Problem (ii) can be solved by approaches using Web video groups for a dynamic database [6,8,10,47,55], these approaches still have the following limitation: Generally, the desired degrees of semantic broadness of topics differ depending on each user [23]. However, these methods are based on flat clustering and thus may provide Web video groups including topics with semantic broadness that do not correspond to the user's desired videos.…”

“…In this paper, the hierarchical structure denotes the property of Web video groups being divided into sub-groups. Several methods have been proposed for extracting the hierarchical structure [20,21,23,24,45,46,48,51] and for tracking topic evolution of Web video groups [6,8,10,47,55]; however, our novel method enables simultaneous realization of extraction and tracking of topic evolution in the hierarchical way. For each time stamp, our novel method extracts the hierarchical structure and salient keywords that represent contents of each Web video group on the basis of network analysis [4] using multimodal features, i.e., features of video contents and metadata.…”

“…In previous papers [6,8,10,47,55], retrieval methods that focus on a dynamic database containing Web videos collected at multiple time stamps have been proposed. These methods extract Web video groups, i.e., Web video sets with similar topics at each time stamp, and track changes in contents of Web video groups over time.…”