A multimedia document browser based on multilayer networks
- PDF / 6,146,769 Bytes
- 38 Pages / 439.642 x 666.49 pts Page_size
- 22 Downloads / 172 Views
A multimedia document browser based on multilayer networks Benjamin Renoust1,2 Shin’ichi Satoh2
· Haolin Ren3,4 · Guy Melanc¸on4 · Marie-Luce Viaud3 ·
Received: 3 May 2019 / Revised: 19 June 2020 / Accepted: 15 September 2020 / © Springer Science+Business Media, LLC, part of Springer Nature 2020
Abstract Querying and retrieving relevant information still remains a difficult task, one with a relatively high cognitive cost for users, who usually focus only on the first few pages of results. This issue drives effort to support the exploration of search results through clustering and visualization. This paper contributes to this challenge by providing a visual analytics system that is designed to support search tasks in multimedia document archives. The system provides complex querying, semantic overviews of time, and visual, and textual concepts combined with analysis. All search tasks are supported with linked-highlighting and leapfrog interactions. This is made possible all in a single data structure thanks to multilayer network modelling. Keywords Multimedia analytics · Visual analytics · Search · Browser · Multilayer networks
1 Introduction Querying and retrieving relevant information still remains a difficult task, one with a relatively high cognitive cost for users [5]. While relevance ranking (such as page ranking [6, 12, 60] based on network modeling) is a powerful approach to extract a set of interesting pages, studies have shown that users usually focus on the first few pages of results [73, 74] (a behavior that barely changed over time [72]). Semantic ambiguity remains, challenging the information retrieval workflow, from extraction down to restitution to users [38, 55], Benjamin Renoust
[email protected] Haolin Ren
[email protected] 1
Institute for Datability Science (IDS), Osaka University, Osaka, Japan
2
National Institute of Informatics (NII), Tokyo, Japan
3
French National Audiovisual Institute (INA), Paris, France
4
LaBRI CNRS UMR 5800, University of Bordeaux, Bordeaux, France
Multimedia Tools and Applications
for which context is key to disambiguation. This has been the focus of the Search Result Clustering (SRC) process in its last two steps: labeling and visualization [14]. This has stimulated the study of strategies adopted by users to find their way into the information space [46]. Information visualization soon responded to tackle this task [2, 80]. In particular, tag clouds can give an overview of this semantic space and support exploratory tasks [70]. The visual inspection of results and even snippets of the first few most relevant results cannot allow users to build a proper mental map of this space. Visual analytics of multimedia data (or multimedia analytics [15]) proposes to extract high-level representations and support high-end analysis of multimedia concepts. Not limited to the textual information, the analysis of video content itself matters, due to the impact of images to the viewers [7]. Advance in computer vision now allows the extraction of visual semantic con
Data Loading...
