1. CiteRivers: Visual Analytics of
Citation Patterns
Florian Heimerl, Qi Han, Steffen Koch, Thomas Ertl
VAST 2015, Chicago, Illinois, USA
20 Aug. 2015
CS 725/825 - Information Visualization
Presented by: Kamlakant
11/10/2017
2. CiteRivers
New visual analytics approach that enables users to make
sense of publication sets by better understanding the
dynamics within a scientific field.
• Community structure of neighboring field
• Popularity of publication
• Most prolific author
• Information combined with the full document
2
3. Motivation
All other previous approaches in this context are based on
• Pure data mining methods
• Static visualization
Examples:
1. IN-SPIRE:
Text analytics tool to identify and track research topics and their
development over time.
Limitation: Does not support citation analysis.
2. Eigenfactor project:
based on citation network analysis to identify important journals across
disciplines and analyze their citation links.
Limitation: Does not take the publication contents into account.
3
4. CiteRivers: A unique approach
• Facilitate joint
analysis of contents
and citations
• Features an
extended version of
a stream graph to
depict clusters over
time.
• Citation word cloud,
trend, diversity,
author and
publisher venue.
4
6. A. Document Panel (Stream graph)
According to the attributes or similarities of the documents, the
documents are divided into a cluster (flow); each cluster (flow) is
composed of blocks of different time; the corresponding keyword
cloud is extracted from the papers in each block.
• Different color coding different clusters (flow)
• The height of the block encodes the number of
corresponding documents
• The font size of the key word coding
• Hover: Highlight the corresponding cluster (stream)
(increase the saturation) and block (transform the background color)
B. Document Clustering Level Slider:
Controls the number of clusters (streams) 6
7. C. Citation aggregation panel
(reference property)
Shows the statistical properties of references to documents in
each block of the highlighted stream, including: Reference age
and reference entropy.
• The age of the cited document refers to the time from the
publication of the cited document to the reference.
• Reference entropy, reflecting the breadth of the reference and its
changes, obtained through the formula.
7
8. D. Author panel
In each block of the highlighted stream,
the top 10 authors were counted and the
changes were shown.
• A total of 12 colors, the same piece,
different color coding of different authors
• The size of the circle encodes the author's
yield, sorted in descending order of yield for
each block
• Hover: Show author's name
• Recognition of same author 8
9. E. Citation flow panel (References)
Via Flow graph method each entry shows:
• Journal / Magazine, publication time, number of citations.
Clustered by the number of references in descending
order
• Double-click: Open the journal / magazine in the
corresponding page in DBLP
F. Categories slider is a reference Cluster
Level Slider
Controls the number of clusters
9
10. G. Document trend plot (literature trends)
Through the scatter diagram, showing the success of
the literature and novelty.
• Achievement, number of citations, value of response
documents and recognition
• Circle brush selection: Click the blank space to display
the author, the article title
• Selected: (1) shows the author, the article title (2)
shows the relevant articles before and after
(normalized) and the clustering of their respective
documents (background color tips)
10
11. Use case analysis
11
1. Literature panels clustered the literature into three categories.
2. VAST was started in 2006, so purple clusters have data since 2006.
3. In Figure a, the blue CitationAge curve, a peak that appeared in
2006, followed by a steady decline in 2007/2008 and a gradual
easing after 2009.
SciVis
InfoVis
VAST
12. Findings: Citation behavior within VAST
12
Citation is heavily concentrated in the
field of “visualization and data
mining”
In this regard, the authors expressed great
interest, so they reviewed the references to
data mining over many years (Figure 6) and
their own references to VAST (Figure 7). It can
be seen that the overall citations are on the
rise. In addition, there is no data from 2005
onwards in Figure 7, again due to the fact that
VAST started in 2006.
13. Conclusion……
13
CiteRivers is an effective approach for gaining insight into the
thematic dynamics of a scientific field, and their relation to
other communities through their citations.
14. ….. Conclusion
14
Insufficiencies:
• Minor bug in document editing, poor quality of chart (missing time)
• Did not reflect the author's cooperation
• Not reflected author unit information
• Did not reflect the reference network
• To explain the document similarity
• The document panel does not support zooming
Future work:
• Author Cooperation and Reference Network
• Compatible with bottom-up analysis
• Compatible with patent documents
Editor's Notes
This visual tool is aim to help explore the citation network of the given publications (conference proceeding). It shows the citation word cloud, trend, diversity, author and publisher venue.
Entropy: The more citation clusters (see: (e) the Citation Panel), the more uneven the distribution, the greater the entropy.
Recognition of the same: (1) neighboring block: direct connection; (2) non-neighboring blocks: a circle in the left / right projections increase, suggesting the same authors also left / right appears, displaying connection hover
Selected: (1) (a) In each block of the document panel, the number of selected authors is displayed (2) (g) Updated in selected literature trends
F. Selected: (1) (a) Number of documents referenced in "Selected Journals / Magazines" in each block of the documentation panel (2) (g) Updated literature references to Selected Journals / Magazines in literature trends