CiteWiz: A Citation Network Visualizer

Niklas Elmqvist∗ Philippas Tsigas†

Department of Computing Science Chalmers University of Technology SE-412 96 G¨oteborg, Sweden

ABSTRACT paper has a dependency to the cited paper. Admittedly, mutual cita- tions cannot be represented and must be either removed entirely or This short paper describes CiteWiz, our tool for citation network broken arbitrarily. . The tool includes a static timeline to present the causality and importance of articles and papers in a citation 3 THE CITEWIZ PLATFORM database. Furthermore, we claim that citations in scientific arti- cles can be modelled as causal relations, and go on to present an The CiteWiz platform uses an open and extensible architecture adaptation of the Growing Polygons technique that makes use of to allow new visualizations to be implemented and added to the the strengths of the method to visualize influences and citations in tool. The platform currently contains two visualizations; a Grow- a set of articles. ing Polygons implementation, and a static timeline diagram. CR Categories: H.5.2 [Information Systems]: User Interfaces; I.3 [Computer Methodologies]: 3.1 Growing Polygons Keywords: citation networks, bibliographic visualization, infor- Views built by the user form the input for the visualization tech- mation visualization, causal relations niques supported by CiteWiz. As mentioned above, the primary vi- sualization technique is currently the Growing Polygons [1] method 1 INTRODUCTION for visualization of general causal relations, suitably modified to be able to handle citation networks and the scalability issues associ- We present CiteWiz, an extensible platform for citation network ated with these. We believe the focus on influence and causality visualization in scientific articles. CiteWiz uses a central citation visualization in the Growing Polygons technique makes it very well database as input to a set of included visualizations: an adapta- suited to citation networks (see Figure 1 for a screenshot). tion of the Growing Polygons [1] visualization technique support- ing large-scale systems of causality, and a new static timeline dia- gram that the citation count and causality of authors or articles to their size and position in time. In this paper, we first present the structure of the CiteWiz plat- form and then go on to describe the analysis process of solving the tasks of the InfoVis 2004 contest [2] using the tool. We close the paper with a discussion of the relative strengths and weaknesses of CiteWiz.

2 CITATIONS AS CAUSAL RELATIONS

A causal ordering is a general relation that relates two events where one is the cause of the other. We can interpret citations in scien- tific articles as causal orderings in at least two different ways: ei- ther with authors as the active entities (processes) and their papers as events, or with papers as the active entities and a single event marking the paper’s publication for each entity. For both cases, we represent citations by causal relations between the events. In our system, we have chosen the latter approach for the simple reason that the former causes problem with the visualization when authors combine to work together on a paper. Seeing that a citation in a scientific article can be modeled by a Figure 1: Partially expanded chronological view of the InfoVis con- causal relation is quite straightforward; a citation implies that (a) ferences. the authors have read the cited paper (and thus, indirectly, that the cited paper existed before the citing paper), and that (b) the citing 3.2 Static Timeline Diagram ∗e-mail: [email protected] †e-mail: [email protected] CiteWiz also contains another visualization informally referred to as a “Newton’s Shoulders” diagram1. This visualization creates a

1So named after Sir Isaac Newton’s famous quote in a letter to Robert Hooke in 1676, “If I have seen further, it is by standing on the shoulders of static, non-interactive timeline of either articles or authors in the field and populate it with the appropriate papers. Then we study the central CiteWiz citation database, displaying each entity as an icon view using the Growing Polygons visualization, focusing on the on the timeline according to their publication date (or the date of transfer of information between the different areas. The strength their first publication, in the case of authors). The surface area of of the GP visualization lies in influence display, allowing us to see each icon is scaled proportionally to the amount of citations the ar- how the areas evolved from the first InfoVis conference in 1995 and ticle or author has received (rounded up so that the icon conforms how they crystalized into the current state of the field. to a uniform grid). The timeline is split up into suitable time units (years or months), and each time segment gets assigned space on 4.3 Task 3: The People in InfoVis the timeline equal to the size of the largest entity in the segment. The icons representing the entities for each time segment are then Continuing with the view of research areas from the previous sec- laid out using a greedy algorithm that places the entities in descend- tion, we add a top-level node for a particular researcher (George ing size within the allocated space on the timeline, always trying to Robertson, for instance) and fill it with the papers associated with minimize the distance to the centerpoint of the diagram (see Fig- that researcher. The Growing Polygons visualization now allows ure 2). us to see the involvement of the researcher in the various research areas, including timing information. Furthermore, the visualization

2003 2003 2002 2002 also indicates paper equalities, i.e. to which of the research areas 2001 2001

M. Scott MarshallGuy Melançon 2000 2000 the researcher’s papers belong. Huub van de WeteringYing−HueyElke Fua A. RundensteinerGraham J. WillsJarke J. VanMartin Wijk Wattenberg

1999 1999

Giuseppe Di BattistaEd H. Chi Ioannis G. TollisIvan HermanRich Gossweiler

Paul Whitney Jim Thomas 1998 1998 When it comes to comparing two different researchers with each

J. MyllymakiG. ChenM. LivnyR. RamakrishnanK. WengerD. DonjerkovicK. BeyerS. LawandeBarry G. Becker

1997 1997

James Pitkow William York other (for instance, Stuart Card and George Robertson), we sim- Brett MilashSeth WidoffMichaelChristian SpenkeWilliam Beilken E. LorensenC. Dunmire M. B. Burks J. A. Senn P. Lucas P. J. Stroffolino C. C. GombergQing−Wen FengA. J. KolojechickT. Munzner AnneKenneth Rose M.Thomas Martin Berlage

1996 1996

Allison WoodruffDavid J. CowperthwaiteStuart Card S. F. Roth M. Pottier K. Pennock D. Lantrip A. Schur J. A. Wise V. Crow J. J. Thomas Marti A.M. Hearst Sheelagh T. CarpendaleF. David Fracchia Erik WistrandM. C. ChuahJohn DillJolly ChenPhillipChristopher Barry G. EdHealey Huai−hsinMichael Chi StonebrakerNahum GershonSougata MukherjeaJ. Carriere R. Kazman Mei C. ChuahPaul BurchardScott Hudson John RiedlAllan R. WilksAlexander AikenDeborahLenwood Hix S.Lyn Heath BartramJames T. Enns ply create a new view of the database with top-level nodes for the 1995 1995

Bob Spence Matthias HemmjeJohn Kolojejchick Daniel A. Keim Clemens KunkelAlexander WillettMatthew O. Ward Lisa Tweedie Pak Chung Wong Peter R. Keller

Hans−Peter Kriegel Ramana Rao Ravinder Bhogal researchers and add the relevant papers to them. The GP visualiza- Benjamin B. Bederson John Lamping M. D. Apperley Jade Goldstein George W. Furnas Ken Fishkin Y. K. Leung David WilliamsMary M. Keller R. Daniel Bergeron 1994 1994 tion now makes explicit which papers, if any, the two researchers David UngarBay−Wei ChangJohn T. StaskoRobert K. FranceRobert Spence Edward A. FoxKellogg S. BoothBen SchneidermanTony D. DeRose John Stasko

Peter Pirolli Maureen C. Stone Eric A. Bier Ken Pier David Fox Scott S. Snibbe Jack D. Mackinlay Oren J. Tversky Ken Perlin William Buxton co-authored, and also the citations between the two authors. We can 1993 1993

Paul Chitson

Christopher Ahlberg naturally extend this to comparing more than two authors simulta- Christopher Williamson James D. Hollan Stephen G. Eick

Matthew Chalmers Steven P. Reiss 1992 1992 neously, by constructing the appropriate view.

Stephen M. CasnerRobert R. Korfhage Finally, an author-based Newton’s Shoulders diagram is also a

Brian Johnson Manojit Sarkar

1991 1991 very powerful visualization for getting an overview of the people in

John F. Hughes John Alan McDonald

S. K. Feiner Joe Mattis Stuart K. Card InfoVis. Bernard DimsdaleClifford Beshers George G. Robertson Jock D. Mackinlay Werner Stuetzle

Steven F. Roth Alfred InselbergSteven K. Feiner Andreas Buja

1990 1990

S. K. Card Andries van Dam G. Robertson J. D. Mackinlay 1989 1989 5 DISCUSSION Marylyn E. McGillRoberto Tamassia Peter Eades David HarelWilliam C. Cleveland

Steven FeinerMarc Levoy Teuvo KohonenFrank,G. Halasz 1988 1988

Richard A. Becker 1987 1987 The strengths of the Growing Polygons visualization lies in the dis- G. W. Furnas Edward R. Tufte Jock Mackinlay

1986 1986 play of causal relations (loosely referred to as “influence” through-

William S. Cleveland

1985 1985 out this paper), but the original technique suffered from some scal-

Marc H. Brown

1984 1984 ability limitations. We have addressed these concerns through the 1983 1983 use of time windows and hierarchical views, but the fact remains James D. Foley

1982 1982 1981 1981

1980 1980 that as the number of simultaneously displayed nodes increases, it

1979 1979

Leslie Lamport 1978 1978 becomes harder to overview the diagram. To combat this, we have 1976 1976

1975 1975

1974 1974 augmented the visualization with color-coded arrows for represent- ing citations, as well as a tree view and a navigation window to simplify the hierarchy management. Figure 2: Newton’s Shoulders diagram of the authors in the IV04 The Newton’s Shoulders diagram gives a useful and interesting contest citation database. view of the authors and articles in a database, but for layout reasons we were forced to relax the causality constraint and limit ourselves to the time constraint. Thus, the results might be slightly inaccurate 4 ANALYSIS PROCESS (in that one researcher may not at all stand on the shoulders of an- other researcher even if they happen to be laid out that way in the This section describes the employment of the CiteWiz tool to solve diagram). the tasks of the InfoVis 2004 contest. 6 CONCLUSION 4.1 Task 1: Static Overview We have presented CiteWiz, a platform for citation network visual- We use the Newton’s Shoulders diagram to create a static history ization that currently includes two separate visualizations (a Grow- overview of the InfoVis field from 1995 to 2002, allowing us to ing Polygons-based influence visualization, and a static timeline di- study the causality and important of authors and articles. Figure 2 agram), and furthermore we have showed how to use the tool to is an author-based timeline for this database (actually, the first en- solve the tasks of the InfoVis 2004 contest. try is from 1974). According to this figure, Stuart Card, George Robertson, Josh Mackinlay, and Ben Shneiderman are the “giants” of InfoVis. A similar diagram can be constructed for the papers in REFERENCES the database. [1] Niklas Elmqvist and Philippas Tsigas. Causality visualization using animated growing polygons. In Proceedings of the IEEE Symposium on 4.2 Task 2: Characterize the Research Areas Information Visualization 2003, pages 189–196, October19–21 2003. [2] Jean-Daniel Fekete, Georges Grinstein, and Catherine Plaisant. In order to characterize the research areas within the InfoVis InfoVis 2004 Contest: The History of InfoVis, 2004. database, we construct a hierarchical view of the areas within the http://www.cs.umd.edu/hcil/iv04contest/. giants.”