Visual Exploration of Hierarchical Data Using Degree-of-Interest Controlled by Eye-Tracking Nelson Silva1,2 , Lin Shao1 , T. Schreck1 , Eva Eggeling1,2 , and Dieter W. Fellner1,3 1 Institute for Computer Graphics and Knowledge Visualization, Graz University of Technology, Austria 2 Fraunhofer Austria Research GmbH, Austria 3 Interactive Graphics Systems Group, Technische Universitaet Darmstadt, and Fraunhofer IGD, Germany 1 Introduction Abstract In this paper, we consider using eye-tracking informa- tion to create an adaptive Visual Analytics system. A E↵ective visual exploration of large data sets main idea of Visual Data Analysis is to support ana- is an important problem. A standard tech- lytic reasoning by interactive visual interfaces to data. nique for mapping large data sets is to use This typically involves the integration of capabilities hierarchical data representations (trees, or of data analysis in terms of visual information explo- dendrograms) that users may navigate. If ration, and the computation capabilities of computers the data sets get large, so do the hierar- to create capable knowledge discovery environments chies, and e↵ective methods for the naviga- [KMSZ06, ABM07]. The need for e↵ective data anal- tion are required. Traditionally, users navi- ysis solutions is obvious as more and more digital infor- gate visual representations using desktop in- mation is being generated and collected in many areas, teraction modalities, including mouse interac- e.g., in medicine, science, education, or business. Data tion. Motivated by recent availability of low- analysis problems can be diverse, such as the amount cost eye-tracker systems, we investigate ap- and speed of data being generated, while it needs to plication possibilities to use eye-tracking for be processed and analyzed. Other problems are re- controlling the visual-interactive data explo- lated with the filtering, aggregation and visualization ration process. We implemented a proof-of- of this same data. concept system for visual exploration of hier- There are di↵erent types of data, among which archic data, exemplified by scatter plot dia- graphs and networks are important data structures to grams which are to be explored for grouping model many relevant data sets [vLKS+ 11, HMM00]. and similarity relationships. The exploration The sizes of graphs grow quickly in many domains, includes usage of degree-of-interest based dis- and these sizes hinder visual exploration of the data, tortion controlled by user attention read from as visualization of large graphs is a challenge. As the eye-movement behavior. We present the basic above cited surveys show, there are numerous graph elements of our system, and give an illustra- visualization methods available, however displaying tive use case discussion, outlining the applica- just a few thousands of nodes e↵ectively remains a tion possibilities. We also identify interesting problem. Therefore, data reduction, e.g., by cluster- future developments based on the given data ing/collapsing of graph nodes is a common approach views and captured eye-tracking information. to limit the data complexity. In terms of hierarchies (trees, or dendrograms) these can be reduced to show Copyright c by the paper’s authors. Copying permitted for only a certain depth of the hierarchy, and group to- private and academic purposes. gether all elements of a sub tree at the sub tree root. In: W. Aigner, G. Schmiedl, K. Blumenstein, M. Zeppelzauer (eds.): Proceedings of the 9th Forum Media Technology 2016, Traditionally, we can use pointing devices to zoom, St. Pölten, Austria, 24-11-2016, published at http://ceur-ws.org pan and navigate to other areas of the graph, ex- 82 Visual Exploration of Hierarchical Data Using Degree-of-Interest Controlled by Eye-Tracking pand/collapse nodes, or adjust the level of abstrac- studies for capturing user eye movements as an input tion [GST13]. However, several problems may arise mechanism to drive system interaction. with the mouse navigation, zoom and expand collapse Etemadpour et al. [EOL14] address eye-movement strategies. When the user pans a graph, the mouse tracking on user studies regarding the accomplishment click can be anywhere on the graph, including empty of typical analysis tasks for projected multidimensional parts of it; also, when the user stops panning, the data, such as tasks that involve detecting and correlat- mouse cursor is “parked“ somewhere in the visualiza- ing clusters. The authors examine and draw conclu- tion, and no useful information can be inferred regard- sions on how layout techniques produce certain char- ing the user intents. acteristics that change the visual attention pattern. Existing eye-tracking devices allow to track the fix- In lab-based user experiments using eye-movements ation areas of a user in front of a display. Among tracking, large and complex gaze trajectory data sets others, eye-trackers are often used for evaluation pur- are generated. There is work which develops tools poses, or to experimentally study human visual atten- to help understand eye-movement patterns [BKR+ 14]. tion. In our work we are interested in the question, These should support the definition and exploration if eye-tracking information can benefit the visual data of a large number of areas-of-interest (AOIs). Eye- exploration process, in addition or as an alternative to movement tracking data is usually analyzed using dif- standard interaction approaches. Nowadays, we can ferent methods [HNA+ 11] and visualization techniques use a↵ordable eye-tracker devices [SZCJ16], like the [BKR+ 14]. Our work follows an AOI-based approach. EyeTribe system1 to monitor the behaviors of the users One important question refers to the justification while exploring a visualization. According to our ex- of why should one use eye-movement tracking and not perience, the device allows a useful tracking of the user just the typical pointing devices, such as, the mouse. gazing on specific regions and individual nodes, on a Many past studies debate the correlation between eye- comparably large tree to be explored. movement tracking and mouse movements. These We present a concept and preliminary implementa- studies presented values from as high as 84% (in a tion of an approach to apply eye-tracking for the pur- study from 2001) [CAS01], to 69% [Coo06], to as low pose of supporting visual exploration of large graphs. as 32% [RFAS08] (in a study from 2008) of correlation Our assumption is that the user gaze indicates areas between eye and mouse movements. These results are of interest in a tree, and consequently we can use this usually dependent on the design of the user interfaces. information to dynamically expand or compress parts Another relevant discussion is centered on the many of the tree. Furthermore, we can also capture a visual advantages of using eye-movement tracking analysis history of the exploration process during which a user and on how to perform a correct eye-movement track- explores a tree view, with applications for example of ing evaluation [JK03]. Eye-movement tracking allows ‘replaying’ analysis sessions or documenting interest- for a fast and continuous tracking of the interest of the ing findings done along the way. This paper is our first users in real time, allowing the detection of moments of step towards an experimental system by which we can confusion, indecision and high interest regions [GW03]. explore design alternatives for eye-based interaction Also, previous studies discuss an important link be- and visualization, as well as to conduct user studies. tween cognitive processes and eye-movements [Hay04]. The accuracy of eye-movement tracking can be kept high by designing a user interface where the size of 2 Related Work the areas of interest is big enough and in accordance We briefly provide an overview of possible applications to the eye-tracker characteristics and the experiment of eye tracking in evaluation, and as an interaction setup. More and more, new eye-trackers are also less modality. We also discuss visualization of hierarchic a↵ected by negative technical factors, e.g., the users’ data and degree-of-interest techniques. head movements that usually reduce the accuracy of the eye-tracker device and calibration difficulties. 2.1 Eye-Tracking and Applications 2.2 Hierarchy Visualization and Focus-and- Eye-movement tracking is a method that is used to Context study, among others, usability issues in Human Com- puter Interaction (HCI) contexts. Pool and Ball In this work we consider visualization of hierarchic [PB05] give an introduction to the basics of eye- data. While hierarchies arise in many contexts, one movement technology, and present key aspects and prominent use of hierarchies is in data clustering. Gen- metrics of practical guidance in usability-evaluation erally, hierarchical graphs together with one of the many clustering techniques [MRS08] can form bene- 1 https://theeyetribe.com (accessed 09/2016) ficial tools for the visual exploration of large data sets. 83 Visual Exploration of Hierarchical Data Using Degree-of-Interest Controlled by Eye-Tracking This visual exploration can be done in the form of Input to the clustering is a distance matrix between trees (or dendrograms), due to their potential for vi- the set of scatter plots. The latter is obtained mak- sual abstraction [CdART12]. In a hierarchical cluster- ing use of image features, which have been shown to ing, users may chose the level of detail by which they work well for the comparison of scatter plots [SvLS12]. explore data. Areas more close to the root contain More precisely, we compute a 25-dimensional intensity more aggregate information, and areas closer to the histogram for each plot. Then, we use the Euclidean leafs include more detail data. distance between histograms to compute the distance An e↵ective interaction technique for navigating (average linkage) of each plot. Using these visual fea- large visualization spaces is to control the level of de- tures, the scatter plots can be arranged hierarchically tail information shown throughout a given visualiza- (e.g., in a tree or circular layout) and the exploration tion. Furnas [Fur86] defines a degree of interest func- for visually similar plots becomes more efficient. tion (DOI-function) where to each node in the graph structure an interest score is defined. This score in 3.2 Hierarchical Layout of Scatter Plots turn is used to expand important areas while reduc- Typically, there are a large number of scatter plot ing other less important areas. Lamping et al. [LR96] views for a high-dimensional data set, these views demonstrated a focus+context (fisheye) scheme for vi- grows quadratically with the number of data dimen- sualizing and manipulating large hierarchies. Gener- sions. Specifically, an n-dimensional numeric data set ally, the expansion or reduction can operate on di↵er- can be represented in n⇥(n2 1) distinctive views us- ent aspects, e.g., on the geometric, semantic or data- ing two distinctive dimensions. To facilitate the ex- oriented level. Previous work [PGB02] was done re- ploration, we take the computed feature vectors of garding the dynamic re-scaling of branches of the tree the scatter plots and apply a hierarchical clustering to best fit the available screen space with an opti- to structure the plots based on their visual similari- mized camera movement. Concerning the aspect of ties. Thus, we receive a structured representation of developing adaptive visualizations, an important sur- the space of scatter plots that arranges similar scatter vey [CK15] was presented that highlights many tech- plots spatially close. To create the hierarchy structure, niques for emotion-driven detection, measurement and we compute the average distance (average linkage) of adaptation, among others. These are very relevant for each scatter plot and build a dendrogram tree, which our concept of adaptation based on degree-of-interest. contains all scatter plots on the leaf node level, see Figure 3. As usual, the dendrogram height describes 3 Concept for Visual Exploration of the similarity (histogram distance) of the scatter plots. Hierarchical Data Guided by Eye Tracking 3.3 Degree of Interest for Navigation of the Hierarchy Next, we introduce our concept (Figure 1) for ex- ploring hierarchical data based on degree-of-interest, The above described dendrogram provides a useful guided by eye tracking. We will exemplify our concept spatial organization of the input space (scatter plots). by using scatter plots as the leaf elements in the hier- Yet, the tree may still be large and complex, especially archy, which are to be explored by a user. The hier- if we have a large number of leaf and internal nodes archy is created by a hierarchical clustering algorithm to inspect and compare. Hence, we introduce spatial using feature-based similarity between scatter plots. distortion to enlarge parts of the tree currently being Our approach relies on eye-tracking to determine the looked at by the user, while visually aggregating the degree-of-interest, which in turn distorts (i.e., magni- remainder of the tree. To this end, we apply eye track- fies/compresses) the hierarchy display. ing using an EyeTribe (see Section 1) setup to track user gazes. Specifically, we measure the user atten- tion on the tree nodes to compute a degree-of-interest 3.1 Considered Application: Hierarchy of (DOI) score for the elements of the dendrogram. Ini- Scatter Plots tially, we show the overall dendrogram using semantic For the exploration of complex data sets, target visu- zoom to fit the whole hierarchy onto the display space. alization techniques such as scatter plots, parallel co- From there, the user starts the graph exploration from ordinates or glyph representations can be used to dis- any point in the view space. While the user navigates cover interesting findings in the data. In our approach, through the view space, the eye tracker captures the we rely on a set of scatter plot visualizations to rep- gaze path. When the user explores specific branches resent all pairwise combinations of a high-dimensional of the tree or local nodes, the eye gaze path and eye- data set. To explore a potentially large set of scat- fixation durations are recorded for each link and node ter plots hierarchically, we apply hierarchic clustering. of the tree. Therefore, besides a measure of interest 84 Visual Exploration of Hierarchical Data Using Degree-of-Interest Controlled by Eye-Tracking based on similarity between scatter plots, we can now overview of dendrogram areas visited so far (see Figure update interest metrics based on time and number of 2 for a gaze history view). This view allows to keep visits to a node. track of visited and unvisited areas, and constitutes Potentially, this recording can also be done for lo- input for further data analysis (see also Section 5). cal parts of the scatter plots, i.e., tracking if the user is dedicating more viewing time to certain local areas in a plot. Such analysis may be useful to detect e.g., correlations, dense areas or clusters in a given plot. Each scatter plot involves the representation of vari- ables (for x and y axes respectively), the interest of the user on these variables (axis) can also be tracked. In the next section, we apply the current eye gaze loca- tion in order to focus the display using semantic zoom. Conceptually, more applications are possible (see also Figure 2: Gaze History Mock-up: It can be activated Section 5). in the navigation panel. The user can track tree areas explored so far by an overlaid trace path (red line). It 3.4 Degree-of-Interest Visualization Using serves as a global map of explored/unexplored areas, Eye-Tracking and it is used for further analysis (see Section 5). We apply eye-movement information to allow the user to navigate through a hierarchy of clusters of scatter plots using semantic zoom. We define an eye-tracking 3.5 Benefits of Our Approach mode, which if enabled, controls the expansion and col- We did informal, preliminary tests of our proposed lapsing of sub-trees in the display based on eye fixa- navigation with 10 users. The feedback so far was pos- tion. Specifically, the area where the user looks at is vi- itive, both to the semantic zoom mode and the gaze sually expanded, revealing the scatter plots under the history view. The navigation was considered as rather sub-tree. The neighboring (remaining) sub-trees are smooth, and users can navigate without larger difficul- represented using just node and link symbols. While ties. Just by looking slightly away in the tree view, the they do not show particular scatter plots, this reduced corresponding movement is initiated in a very intuitive representation is still indicating basic data properties way. This facilitates the entire process of exploring the like number of scatter plots represented, or structure data in the tree, i.e., the view panning is synchronized of the similarity relationships within the dendrogram. with the field of view and the eye-movements of the user. When the user stops using the eye-based nav- igation, attention information starts to be collected again (eye-gaze duration on each area-of-interest) and it is the basis for the analysis of user interest detection and possible subsequent recommendation of interest- ing views. Note that in our concept we consider only gaze-based navigation. Of course, we can rely in addi- tion on mouse/keyboard input to facilitate navigation, e.g., for labeling, saving views/bookmarking, etc. 4 Implementation and Application To test our approach, we developed a proof-of-concept Figure 1: Concept: Exploration of large scatter plot allowing the exploration of a large tree (dendrogram). spaces. The user eye-gaze is detected, leading to an expansion of the focused sub-tree (center rectangle). 4.1 System Implementation The remaining data is shown using a node-link repre- In our tree visualization, the leaf nodes are composed sentation (context, outside center rectangle). of visual representations of the data, i.e., scatter plots with pairs of data attributes. We use our own mod- When the user stops the eye-tracking mode, the ap- ified version of the JUNG system [OFWB03] for the plication goes back to a state where it tracks only the tree visualization. We made changes on the adjust- user interest on each specific node, i.e. eye-gaze dura- ment of the lens size in the view space and position- tion on each node (no pan control). We also show an ing of the lens, now they are controlled by user eye- 85 Visual Exploration of Hierarchical Data Using Degree-of-Interest Controlled by Eye-Tracking movements and updated in real-time. We created a 4.2 Application customized tree layout to display color-coded nodes according to the computed similarity distance measure Our data set is retrieved from the Eurostat 3 data (darker color = lower similarity), and also the ability repository, which provides a collection of data sets to display visual representations of the data on the leaf containing information on EU related topics (e.g., nodes, i.e, scatter plots. economy, population and industry). We use a pre- The initial preset for DOI specification is the calcu- processed data set from preliminary work [SSB+ 15], lated similarity distance between scatter plot images. which contains 27 statistical attributes from 28 EU For this calculation, we make usage of a basic descrip- countries showing temporal changes over time. tor from the Java Image Processing Cookbook 2 that is All navigation actions presented in the following ex- based on the average calculation of 25 color triples for amples illustrate a typical usage of our navigation sys- each image. After performing the comparison between tem. Figure 4 shows an example of an ideal view over each image using the descriptors, we create a distance a small data portion, where the user is able to see the matrix with the computed distances between all im- majority of the data. Practically, for larger data sets ages. This matrix is handed out to an agglomerative users will often have a more narrow view over the en- hierarchical clustering algorithm [Beh16]. tire tree, depicted in Figure 5 with the 3 narrow views. We tested our system with several hierarchical trees. Here, we illustrate the application of a hierarchical tree exploration of our data set. At the root and top sub- trees we can find information about clusters of similar scatter plots, and at the leafs we find individual scat- ter plots. For this proof-of-concept we use a dendro- gram comprising 269 scatter plots (leaf nodes) and 536 edges. Figure 3 shows a zoomed in view of the tree, the color-coded nodes according to distance similarity of the scatter plots, and the circular zooming lens (in gray color). In the navigation panel (top-right corner), we can get an overview of the entire tree size and re- spective available view space. The current view size Figure 4: Ideal Case: Users can view several clusters (depending on the zoom level) and location is denoted of related scatter plots at once. Due to limitations of by a white rectangle. The lens can be used to per- display space, this is often not possible, hence the need form a close zoom into the scatter plot image, and it for adaptive visualization for navigation. can be used to activate the display of a di↵erent visual representation of the data. The navigation order (view sequence) followed by the user on these narrowed views can be random, it might just follow the similarity distance measures (de- picted by the color-coded node rectangles). Figure 5, shows that the user first moved to view V1, where a group of interesting clustered scatter plots is visible (Figure 6). In view V1, the system detects a high gaze duration and infers that the interest is on one of the scatter plots (marked with ”*”). After an in-depth inspection of this area, the user navigates to a view V2 (Figure 7) over another group of clustered scatter plots. The next most interesting and similar scatter plot (yellow color) is occluded in view V2 and it is Figure 3: Zoomed-in view of the hierarchical tree. The only visible in view V3 (Figure 8). mouse wheel can be used to: increase/decrease the lens It might take time until an interesting scatter plot magnifying ratio; increase/decrease the size of the cir- (view V3) is spotted by the user. Also, there might cular zooming lens (gray circle) by clicking and drag- exist other interesting scatter plots in another part of ging its border. A navigation panel (top-right corner) the tree, in a more far, and yet hidden location, e.g., gives an overview of the actual position in the tree. Figure 9. 3 Statistical Office of the European Union (http://ec. 2 Java Image Processing Cookbook (http://goo.gl/FBXbjp) europa.eu/eurostat). Accessed 09/2016. 86 Visual Exploration of Hierarchical Data Using Degree-of-Interest Controlled by Eye-Tracking Figure 5: In practice, users may have limited views Figure 7: View 2: The user moves to a new location, over a large space that must be explored. Therefore, but misses an interesting scatter plot that is occluded the views (V1, V2, V3) might be limiting and not fol- on a top location. lowing an ideal sequence of exploration that would lead to finding interesting factors and to the creation of a useful mental model while exploring the data set. We take the eye-gaze duration time in account to infer about the interest of the user. Figure 8: View 3: User navigates to this location and finds an interesting scatter plot related to view V1. Figure 6: View 1: Realistic view of a first set of scatter information presented, but also to capture longer se- plots. The user is focused on a set of scatter plots and quences of visual exploration. The analysis of such unaware of other interesting locations of the tree. captured data presents manifold opportunities to en- hance the analysis process. For example, similar to previous work on navigation recommendations for ex- In summary, our example merely demonstrates ploring hierarchical graphs [GST13], approaches could some of the challenges associated with the exploration be developed to suggest what new parts of the tree of large graphs. The duration of the gazes can be used should be explored next by the user. to expand or collapse sub-trees and hence provide a more organized (less cluttered) overview of the data, For now, our user focus model considers all elements reducing the risk of getting lost in the exploration of of the tree (inner nodes and scatter plots). Given suf- large dendrogram trees. However, our measures com- ficient tracking resolution, we may apply the degree- puted directly from the location of the gazes are only of-interest concept also locally within a focused scatter a first step to control the views. We plan to collect plot. There are numerous ways to heuristically com- data about the eye-movement scan paths and respec- pute interest measures from eye-gaze fixations, fixation tive eye-gaze durations, as well as recurrences to de- sequences, and gaze recurrences. Examples include velop more adaptive hierarchy views. learning relevance of local patterns, or deducing data groups of interest to a given user or analysis session. In the future, we hope to leverage such information 5 Discussion and Extensions and detect important aspects of the analysis problem We implemented a proof-of-concept system for which at hand (e.g., whether there is exploratory or confir- we see numerous extension possibilities. First, our so- matory analysis going on in a given session), or detect lution allows not only to adjust the amount of visual the level of user expertise. Depending on this infor- 87 Visual Exploration of Hierarchical Data Using Degree-of-Interest Controlled by Eye-Tracking 6 Conclusion We presented a concept for visual exploration of hier- archically organized data, that relies on eye-tracking to steer the level of resolution shown. We assume that long gaze fixation times indicates user interest and hence can be used as a proxy to control the visual display of large data. Our e↵ort extends previous work by a new user interface, allowing the navigation and the setting of the degree-of-interest to be determined by eye-movements, and it can be applied on both desk- top screens or larger displays (e.g., using wearable eye- trackers). We applied this idea to the specific problem of comparing scatter plot diagrams, and hence support Figure 9: View 4: In a more far location (from V1, V2 a type of meta-visualization: the elements in the tree and V3) the user notices that there is another inter- are complex objects (visualizations). To this end we esting scatter plot worth investigation. applied dendrogram computation based on image fea- tures, an approach which may help to overview large amounts of data views by grouping these for similar- mation, the system may adapt its presentation and ity. We have shown illustrative use cases for how eye- functionality accordingly. Also, a view recommender tracking can enhance a hierarchical data visualization, module may prevent the user from repetitively going by mapping eye-gazes to degree-of-interest represen- to already examined areas in the display, suggesting in- tations. Yet, our work is in an early stage and we stead, previously unseen parts of the view space, based see ample areas for future work. Future work includes on analysis of the eye-movement (scan) path. To that high gaze-tracking precision on each node, refinement end, it is interesting to ask how one can do sugges- of interaction operations, view recommending, adap- tions of other scatter plots to be explored, e.g., based tive visual representations, and analysis provenance vi- on visual or data-driven similarities. sualization. Finally, evaluation of our approach should Another idea is to choose adaptively di↵erent visual be done in comparison to non-eye-tracking controls representations on the nodes based on advanced inter- to qualitatively or quantitatively assess strengths and est measure computation. For example, in our current weaknesses of the approach. visualization approach (scatter plots) we might want to change dynamically between scatter plots, table rep- References resentation, parallel coordinates visualization, or re- [ABM07] Wolfgang Aigner, Alessio Bertone, and Silvia gression or clusters models computed for a given scat- Miksch. Tutorial: Introduction to Visual Ana- ter plot. Also, an interest function should be adaptive lytics, pages 453–456. Springer Berlin Heidel- also regarding time, e.g., during a longer or repeated berg, Berlin, Heidelberg, 2007. analysis cycle. Such interest functions might take into [Beh16] Lars Behnke. Implementation of an agglomer- account di↵erent objectives, e.g., the user might want ative hierarchical clustering algorithm in java., to explore the most dissimilar clusters of scatter plots, 09 2016. or explore all scatter plots that have a certain shape. Depending on these objectives the interest function [BKR+ 14] T. Blascheck, K. Kurzhals, M. Raschke, M. Burch, D. Weiskopf, and T. Ertl. State- might need to be adjusted. of-the-Art of Visualization for Eye Tracking We also mention that our gaze path visualization Data. In R. Borgo, R. Maciejewski, and I. Vi- could be enhanced to work as an overview tool to ola, editors, EuroVis - STARs. The Euro- represent explored/unexplored regions. A gaze path graphics Association, 2014. might eventually serve as a visual history of a whole [CAS01] Mon Chu Chen, John R. Anderson, and exploration process. Therefore, we may extend a given Myeong Ho Sohn. What can a mouse cur- gaze path by annotating certain views visited (e.g., sor tell us more?: Correlation of eye/mouse scatter plot thumbnails) at certain points in the gaze movements on web browsing. In CHI ’01 Ex- path (e.g., exceptionally long or short fixation times). tended Abstracts on Human Factors in Com- Appropriate visual design might communicate a whole puting Systems, CHI EA ’01, pages 281–282, analysis session in a single image, which would be a New York, NY, USA, 2001. ACM. valuable tool for reproducibility and communication [CdART12] Stéphan Clémençon, Héctor de Arazoza, Fab- of analysis sessions. rice Rossi, and Viet-Chi Tran. Hierarchi- 88 Visual Exploration of Hierarchical Data Using Degree-of-Interest Controlled by Eye-Tracking cal clustering for graph visualization. CoRR, [LR96] John Lamping and Ramana Rao. Visualizing abs/1210.5693, 2012. large trees using the hyperbolic browser. In [CK15] Daniel Cernea and Andreas Kerren. A sur- Conference Companion on Human Factors in vey of technologies on the rise for emotion- Computing Systems, CHI ’96, pages 388–389, enhanced interaction. Journal of Visual Lan- New York, NY, USA, 1996. ACM. guages and Computing, 31(Part A):70–86, [MRS08] Christopher D. Manning, Prabhakar Ragha- 2015. van, and Hinrich Schütze. Introduction to [Coo06] Lynne Cooke. Is the mouse a” poor man’s Information Retrieval. Cambridge University eye tracker”? Annual Conference-Society for Press, New York, NY, USA, 2008. Technical Communication, 53:252, 2006. [OFWB03] J. O’Madadhain, D. Fisher, S. White, and [EOL14] R. Etemadpour, B. Olk, and L. Linsen. Eye- Y. Boey. The JUNG (Java Universal Net- tracking investigation during visual analysis of work/Graph) Framework. Technical report, projected multidimensional data with 2d scat- UCI-ICS, October 2003. terplots. In Information Visualization Theory [PB05] Alex Poole and Linden J. Ball. Eye track- and Applications (IVAPP), 2014, pages 233– ing in human-computer interaction and us- 246, Jan 2014. ability research: Current status and future. In [Fur86] G. W. Furnas. Generalized fisheye views. In Prospects, Chapter in C. Ghaoui (Ed.): En- Proceedings of the SIGCHI Conference on Hu- cyclopedia of Human-Computer Interaction. man Factors in Computing Systems, CHI ’86, Pennsylvania: Idea Group, Inc, 2005. pages 16–23, New York, USA, 1986. ACM. [PGB02] Catherine Plaisant, Jesse Grosjean, and Ben- [GST13] Stefan Gladisch, Heidrun Schumann, and jamin B. Bederson. Spacetree: Supporting ex- Christian Tominski. Navigation Recommenda- ploration in large node link tree, design evolu- tions for Exploring Hierarchical Graphs, pages tion and empirical evaluation. In Proceedings 36–47. Springer Berlin Heidelberg, Berlin, of the IEEE Symposium on Information Vi- Heidelberg, 2013. sualization (InfoVis’02), pages 57–, Washing- [GW03] Joseph H. Goldberg and Anna M. Wichan- ton, DC, USA, 2002. IEEE Computer Society. sky. Chapter 23 - eye tracking in usabil- [RFAS08] Kerry Rodden, Xin Fu, Anne Aula, and Ian ity evaluation: A practitioner’s guide. In Spiro. Eye-mouse coordination patterns on J. Hyn, R. Radach, and H. Deubel, editors, web search results pages. In CHI ’08 Ex- The Mind’s Eye, pages 493 – 516. North- tended Abstracts on Human Factors in Com- Holland, Amsterdam, 2003. puting Systems, CHI EA ’08, pages 2997–3002, [Hay04] Mary M. Hayhoe. Advances in relating eye New York, NY, USA, 2008. ACM. movements and cognition. Infancy, 6(2):267– [SSB+ 15] Lin Shao, Timo Schleicher, Michael Behrisch, 274, 2004. Tobias Schreck, Ivan Sipiran, and Daniel A. [HMM00] Ivan Herman, Guy Melançon, and M. Scott Keim. Guiding the Exploration of Scatter Marshall. Graph visualization and navigation Plot Data Using Motif-based Interest Mea- in information visualization: A survey. IEEE sures. IEEE Int. Symposium on Big Data Vi- Transactions on Visualization and Computer sual Analytics, September 2015. Graphics, 6(1):24–43, January 2000. [SvLS12] Maximilian Scherer, Tatiana von Landes- [HNA+ 11] Kenneth Holmqvist, Marcus Nystrom, berger, and Tobias Schreck. A Benchmark Richard Andersson, Richard Dewhurst, Hal- for Content-Based Retrieval in Bivariate Data szka Jarodzka, and Joost van de Weijer. Eye Collections. In Proc. Int. Conference on The- Tracking. A comprehensive guide to methods ory and Practice of Digital Libraries, 2012. and measures. Oxford University Press, 2011. [SZCJ16] Popelka Stanislav, Stachon Zdenek, Sasinka [JK03] Robert J.K. Jacob and Keith S. Karn. Com- Cenek, and Dolezalova Jitka. Eyetribe tracker mentary on section 4 - eye tracking in human- data accuracy evaluation and its interconnec- computer interaction and usability research: tion with hypothesis software for cartographic Ready to deliver the promises. In J. Hyn, purposes, 2016. R. Radach, and H. Deubel, editors, The [vLKS+ 11] T. von Landesberger, A. Kuijper, T. Schreck, Mind’s Eye, pages 573 – 605. North-Holland, J. Kohlhammer, J. van Wijk, J.-D. Fekete, Amsterdam, 2003. and D. Fellner. Visual analysis of large [KMSZ06] Daniel A. Keim, Florian Mansmann, Jorn graphs: State-of-the-art and future research Schneidewind, and Hartmut Ziegler. Chal- challenges. Wiley-Blackwell Computer Graph- lenges in visual data analysis. In Proceedings ics Forum, 30 (6):1719–1749, 2011. of the Conference on Information Visualiza- tion, IV ’06, pages 9–16, Washington, DC, USA, 2006. IEEE Computer Society. 89