MPVR: A multi-perspective visual retrieval toolkit for multi-dimensional data - Inria - Institut national de recherche en sciences et technologies du numérique Access content directly
Conference Papers Year : 2013

MPVR: A multi-perspective visual retrieval toolkit for multi-dimensional data

Yi Peng
  • Function : Author
Li Chen
  • Function : Author
Quan Li
  • Function : Author
Zhu Zhu
  • Function : Author
Jie Dong
  • Function : Author
Jun-Hai Yong
  • Function : Author

Abstract

Rapid retrieval of specified items in a large amount of multi-dimensional data has become an urgent and challenging topic for improving the effectiveness of data searching and analysis approaches. In order to help users quickly re-find the desired items, we present a novel multi-perspective visual retrieval method which is totally different from traditional visualizations for searching results, such as list-based techniques. In our prototype, customized dimensions are generated adaptively which enables users to fast approach to the desired targets from multiple perspectives. We also design several display modes to layout the items by introducing degree-of-interest(DOI) ranking, which is implicitly defined by the geometric relationships of different dimensions. The display modes includes the Cluster Mode using importance-driven dimensionality reduction techniques, the Radial Mode diffusing the items into separate directions, the Uniform Mode evolved by Archimedean Spiral which keeps items non-overlapping and the Custom Mode for deep analysis. Moreover, we combine Focus + Context, Visual Cues which presents item properties, Collaborative Filtering and animations to recommend results according to the features users are interested in. Finally, we evaluate our method in different scenarios: text message retrospection, desktop search and web browsing history retrieval. The experiments show that our method is flexible, extensible, and efficient.
Fichier principal
Vignette du fichier
a_e_2013c_ae_ae_ae_ae_.pdf (465.25 Ko) Télécharger le fichier
Origin : Files produced by the author(s)
Loading...

Dates and versions

hal-00920670 , version 1 (19-12-2013)

Identifiers

  • HAL Id : hal-00920670 , version 1

Cite

Yi Peng, Li Chen, Quan Li, Zhu Zhu, Jie Dong, et al.. MPVR: A multi-perspective visual retrieval toolkit for multi-dimensional data. The 12th ACM SIGGRAPH International Conference on Virtual-Reality Continuum and its Applications in Industry (VRCAI 2013), Nov 2013, Hong Kong, China. ⟨hal-00920670⟩
102 View
125 Download

Share

Gmail Facebook X LinkedIn More