|
Multimedia analysis
Newdle (News Wordle)
EventRiver
SIB (Semantic Image Browser)
Graph visualization using multidimensional
visualization approaches
PIGVis (PIxel-oriented Graph Visualization)
High dimensional data visualization
MVE (Multivariate Visual Explanation) - website under construction
VAR (Value and Relation Display) - website under construction
Insight management
ClickToAnnotate - website under construction
Newdle
- Online news visual exploration
Summary: Newdle allows users to visually explore online news collections
through their RSS feeds. It provides
novel visualizations and interactions to
allow users to conduct a rich set of visual analytics tasks, such as browsing
hot topics, investigating hot topics, search relevant keywords, search articles
by keywords, comparing topics and
keywords, and conducting extended reading for
articles of interest. Newdle is a visual analytics approach. Its
visualizations
and interactions are supported by underlying automatic article relation analysis
and article network
construction and analysis. It automatically extracts hot
topics from a large collection through clustering analysis
on the article
network. This project is supported by
the DHS international
program,
NSF
IIS-0946400,
and VACCINE.
Paper: Jing Yang, Dongning Luo, and Yujie Liu: Newdle: Interactive Visual
Exploration of Large Online News Collections. Submitted to CG&A
Screenshots:
1.
Hot topics in the New York Times World News from 12/19/2009 to 1/17/2010.
2.
Hot topics with more details. The original news articles can be accessed by
clicking their titles.
3.
Comparing the keywords "Yeman"
and "Abdulmutallab, Umar Farouk"
4.
Extended reading of the news article displayed in the top row.
EventRiver
- Broadcast News Video Visual Exploration
Summary: EventRiver allows users to interactively explore broadcast news
videos collections. It first uses an
incremental clustering algorithm to capture hot topics in the form of a series
of events. It then intuitively conveys
the semantics of the events within their temporal context via visualization
using a river metaphor. It provides users
with a rich set of interactions for interactively browsing events in the
collection, retrieving events of interest, and
conducting detailed visual analysis. This project is supported by
the DHS international
program, VACCINE,
and SVAC.
Paper:
Dongning Luo, Jing Yang, Milos Krstajic,
Jianping Fan, William Ribarsky, and Daniel Keim: EventRiver: Interactive
Visual Exploration of Constantly Evolving Text Collections. Submitted to TVCG.
Screenshot:
An event overview of CNN news from Aug.
1 to 24, 2006 (29,211 news reports) with less significant events
filtered out. Each icon represents an event. The x-axis is the time axis. The
width of an event along the y-axis
indicates its significance. The events in the same color form an event group
showing a developing topic. Several
representative events of significant topics are annotated using dual-labels,
where labels in white background convey
context information while those in yellow convey unique contents of the events.
SIB -
Semantic Image Browser for Exploring Large Image Collections
Summary: SIB allows user to interactively browse and search large image
collections in which images are tagged
by their contents. The major visualization components of this browser are
Multi-Dimensional Scaling (MDS) based
image layout, the Value and Relation (VaR) display that allows effective high
dimensional visualization without
dimension reduction, and a rich set of interaction tools such as search by
sample images and content relationship
detection.
Paper:
J. Yang, J. Fan, D. Hubball, Y. Gao, H. Luo, W.
Ribarsky and M. Ward. Semantic Image Browser: Bridging Information Visualization
with Automated Intelligent Image Analysis. IEEE Symposium on Visual Analytics
Science and Technology 2006:
191-198 (pdf)
Downloadable:
Source code release for windows
Video
Screenshots:
1.
Overviews:
(a) An MDS image overview of the Corel
collection (1100 images). (b) The rainfall image view of
the collection that shows the correlations between the bottom center image and
other images, which are indicated
by the vertical distances between them. It can be seen that there is a
group of images containing snow and
mountains that are very similar to the focus image. A selection by sample image
is applied in this view. The sample
image is highlighted by pink, and the selected images are highlighted by green.
2.
VAR content view:
(a) and (b) show the Corel collection
(1100 images and 20 contents) and (c) show
the MIT collection (4559 images and 29 contents). In (b), the data items are
reordered by their values in the
sailcloth dimension. A selection has been performed to highlight images with the
sailcloth contents.
PIGVis
- Using pixel-oriented displays to explore large graphs
Summary: PIGVis allows users interactively explore large graphs with high
dimensional node attributes. It is
our preliminary research toward using multi-dimensional visualization techniques
to visualized graphs. It displays
a large graph in a clutter free visualization and allows users to conduct a rich
set of analysis tasks, such as cluster
analysis, outlier analysis, neighborhood exploration, and node attribute
exploration, on large graphs with ease. This
project is supported by NSF
IIS-0946400
and the DHS international program.
Paper:
Jing Yang, Scott Barlowe, Yujie Liu: Pixel-Oriented Graph Visualization: a
Multidimensional Visualization
Approach to Visualizing Large, Multivariate
Graphs. To be submitted.
Screenshots:
1.
A graph with
3,197 nodes and 28,277 edges:
Keywords co-occurring with "Bush, George W(Pres)" in
the
New York Times news articles from 9/7/2001 to 9/15/2001 are highlighted in this
graph by underlines. They are
displayed within the context of their closely related keywords. Except the first
row, each row displays the labels
and expending neighborhoods of a cluster of closely related keywords.
 | 
