Exploring Media Transparency With Multiple Views Alexander Rind,1,2 David Pfahler,1 Christina Niederer,2 and Wolfgang Aigner1,2 1 TU Wien, Austria 2 St. Poelten University of Applied Sciences, Austria This so-called media transparency (MT) dataset is a valuable resource for politically concerned citizens Abstract as well as for data journalists [Aus15, Lor10]. They are interested in exploring the available data indepen- Politically concerned citizens and data jour- dently looking for stories beyond prearranged sum- nalists want to investigate money flows from mary statistics. However, the MT dataset is much government to media, which are documented too large to be browsed line by line. Neither is it suffi- as open government data on ‘media trans- cient to look only at the largest flows of money because parency’. This dataset can be characterized many possible questions of interest focus on changes as a dynamic bipartite network with quantita- over time and the many-to-many relationship between tive flows and a large number of vertices. Cur- legal entities and media [NRA+ 16]. For this purpose rently, there is no adequate visualization ap- it is useful to conceptualize the MT dataset’s money proach for data of this structure. We designed flows as time-dependent attributes on the edges of a a visualization providing coordinated multiple bipartite network. Simple data analysis tools such as views of aggregated attribute values as well as spreadsheets do not adequately support such a data short tables of top sorted vertices that can be structure. explored in detail by linked selection across Interactive visual representations of data [CMS99, multiple views. A derived attribute ‘trend’ Mun14] are a well-suited approach to explore com- allows selection of flows with increasing or de- plex datasets. Many visualization techniques have creasing volume. The design study concludes demonstrated their value in exploring time-oriented with directions for future work. data [AMST11] and network data [BBDW16, HSS15]. However, the combination of time with quantitative 1 Introduction flows in a bipartite networks is still an open challenge for visualization research [NAR15]. Independent news and media are a cornerstone of mod- This paper contributes a visualization design study ern democracy – often called the fourth power. How- [SMM12] for time-oriented quantitative flows in a bi- ever, governmental advertisement and sponsorships partite network. It uses the MT dataset as example could influence news coverage limiting the media’s in- and non-expert users such as citizens and journalists dependence. In Austria, the federal law on Trans- as target audience. After surveying related work in parency in Media Cooperation and Funding [Med15] Section 2 and characterizing the domain problem in makes it mandatory to disclose such flows of money Section 3, we present the justified visualization design from legal entities (e.g., federal ministries, cities, eco- in Section 4. Next, a usage scenario demonstrates the nomic chambers, government-owned companies) to design’s utility in Section 5. The paper concludes with media institutions (e.g., newspaper, TV, radio, on- reflections for future development. line). The Austrian Regulatory Authority for Broad- casting and Telecommunications [RTR] collects these data and makes them publicly available via the Aus- 2 Related Work trian open government data portal [RTR16]. The design space of network visualization has been mapped in some recent state-of-the-art reports: Had- Copyright c by the paper’s authors. Copying permitted for private and academic purposes. lak et al. [HSS15] identified five facets of concern for In: W. Aigner, G. Schmiedl, K. Blumenstein, M. Zeppelzauer visualizing a network: (i) its structure comprised of (eds.): Proceedings of the 9th Forum Media Technology 2016, nodes and edges, (ii) partitions, (iii) the attributes St. Pölten, Austria, 24-11-2016, published at http://ceur-ws.org of nodes and edges, (iv) dynamics, i.e., change over 65 Exploring Media Transparency With Multiple Views Table 1: Raw format of the media transparency (MT) dataset (first four entries) legal entity time law medium money amount Abfallwirtschaft Tirol-Mitte GesmbH Q4/2012 §2 Bezirksblätter Tirol 8,122.32 e Agrarmarkt Austria Marketing GesmbH Q4/2012 §2 Falsta↵ 26,418.00 e Agrarmarkt Austria Marketing GesmbH Q4/2012 §2 Connoisseur Circle 6,142.50 e Agrarmarkt Austria Marketing GesmbH Q4/2012 §2 bz-Wiener Bezirkszeitung 7,031.16 e .. . time, and (v) spatialization such as geographic con- fees. Each quarter, each legal entity is obligated to text of nodes. Beck et al. [BBDW16] addressed make a disclosure for both §2 and §4. Every media in particular visual representations for dynamic net- cooperation involving more than 5,000 e needs to be works such as animation and timeline. Von Landes- included with the recipient’s name and the amount of berger et al. [vLKS+ 11] focused on large networks. money accumulated in the quarter. If a legal entity Niederer et al. [NAR15] surveyed visualization of dy- had no such media cooperation, it still has to submit namic, weighted and directed networks, and thus, data a nil report. of a structure similar to the MT dataset. The MT dataset is published on an open data portal Examples of such related visualization designs are [RTR16] each quarter of a year with data covering the DOSA [vdEvW14], egoSlider [WPZ+ 16], egoLines preceding eight quarters. The raw data are formatted [ZGC+ 16], Graph Comics [BKH+ 16], TimeArcTree as semicolon-separated values in a text file. Table 1 [GBD09], and Visual Adjaceny List [HBW14]. How- shows the five relevant variables: name of the legal ever, none of these approaches explicitly considers the entity, time specified by year and quarter, category bipartite nature of the MT dataset, i.e., that there are of legal background, name of the medium, amount of distinct nodes for legal entities and for media. money (quantitative). Additionally, the raw data con- We could identify only one scholarly work focusing tains a variable that flags nil reports. on visualizing the MT dataset in particular: Niederer et al. [NRA+ 16] investigated the visualization needs 3.1 Data Abstraction of data journalists based on four interviews that were anchored on the MT dataset as exemplary scenario. The MT dataset is comprised of the quarterly money Besides that, there is some press coverage on the data transferred from legal entities to media. We can con- and some articles are accompanied by interactive web ceptualize these data as time-dependent flows in a bi- infographics (e.g., derStandard.at [Ham], Paroli Mag- partite network (Figure 1) [NRA+ 16]. The network’s azin [Lan]). Yet, these infographics present a subset underlying graph is bipartite because its vertices can of the available data that has been aggregated and fil- be divided in two disjoint sets – legal entities and tered to support their articles’ story. Since they allow media – and each edge connects vertices of di↵erent only minimal interactivity, further exploration is not sets. These edges are directed and weighted repre- possible. Furthermore, since 2013 the open source soft- senting the flow of money from legal entities to media. ware project Medientransparenz Austria [SBSV] pro- The network is dynamic both in terms of its struc- vides an interactive online tool that shows the com- ture (vertices and edges can appear or disappear over plete MT dataset. It integrates several visual repre- time) and its quantitative flows (weights changing over sentations giving insight into the data, but its views time) [vLKS+ 11]. The time-oriented aspect of the data require much scrolling and are distributed across mul- can be characterized as instants on a discrete, interval- tiple pages. In addition, changes of money flow over time are not explicitly represented. 3 Background As a fundament for developing a novel visualization design for the MT dataset, we must first understand its background and characterize the domain problem. The law [Med15] regulates three categories of Figure 1: Conceptualizing the MT dataset as time- money flows that need to be disclosed: §2 covers adver- dependent flows in a bipartite network tisement, §4 sponsorships, and §31 ORF programme Figure by [NRA+ 16] used with permission. 66 Exploring Media Transparency With Multiple Views based, linear time domain with the granularities quar- The quantitative values of quarterly flows vary be- ter and year [AMST11]. tween e 5,000 (the minimum to be reported) and This abstract data structure has some benefits over e 1.929.533 (median = e 10,931; average = e 23,444). the raw data’s table structure: The central aspects of the problem domain (legal entities, media, and 3.3 Design Requirements flows) are represented directly as data items, which can have properties from derived data such as aggre- Based on our data analysis described above and the gated money. Network metrics such as in-degree can interviews with data journalists interested in the MT be examined. They can also be manipulated by user dataset as reported by Niederer et al. [NRA+ 16], we interaction. can identify five design requirements that a visualiza- tion design for the MT dataset should fulfill: 3.2 Preprocessing and Analysis of Data Scale R1 Data scalability: The number of vertices for both legal entities and media is relatively large. Be- We perform some preprocessing to achieve a better sides the institutions’ names and their network data basis for our visualization: relations, there are no further data that could be (1) We substitute the original MT dataset with used for clustering vertices. While a majority of data from the Medientransparenz Austria project vertices is only sparsely connected, some central [SBSV], which have two benefits: First, they have vertices have a large number of flows. Likewise, included data for all quarters since the start of the the weights representing amount of money can MT dataset in Q3/2012. Second, they have pre- vary widely within the network. The time dimen- processed the data to clean di↵erent forms of writ- sion adds additional scale. ing the names of media and legal entities. Such R2 Development over time: The data journalists in- inconsistencies could result either from typos or terviewed by Niederer et al. [NRA+ 16] expressed from the organization actually being renamed. particular interest in patterns or abnormalities in (2) Next, we discard nil reports from the data. Even the number and weight of flows over time. though these nil reports make up about 80% of all R3 Data wrangling: For two reasons, users would records, they cannot add any insight to our design need to refine the MT dataset by basic data wran- as they have missing values for media name and gling functionality: First, they can add their im- amount of money. plicit expert knowledge into the analysis. For ex- (3) Finally, we also discard programme fees (legal cat- ample, they could group together the federal min- egory §31) because on the one hand there are only istries run by politicians of the same party. one or two records per quarter and on the other Second, data quality is still not sufficient for some hand their amount is much higher than any other data entries even though data quality measures record. The median §31 amount is about 80 times have been taken by the RTR and the dataset has as much as the highest regular amount. been pre-cleaned by the Medientransparenz Aus- tria project. Table 2 shows some examples based As of summer 2016, the preprocessed MT dataset on media from this dataset containing the string encompasses 36,261 quarterly money flows over 15 “standard”. It should be possible to combine en- quarters (Q3/2012–Q1/2016). So that one quarter has tries with di↵erent forms of writing or di↵erent on average circa 2,400 flows. 34,717 flows (96%) have media (print, online, app) of the same newspaper §2 as legal background and there are 1,544 flows for and to hide entries of poor quality. §4. (30 flows for §31 have been discarded.) R4 Ease of use: The target audience of the MT There are 993 distinct legal entities and 3,813 dataset such as interested citizens or data jour- distinct media. Legal entities have between 1 and nalists will most likely have no expert knowledge 1,782 outgoing flows (median = 8; average = 36.2), if of statistics or visualization. They will access the we count each quarter as a separate flow. These flows MT visualization as a spontaneous activity where connect them to between 1 and 618 distinct media (me- no special training can be provided. Therefore, dian = 3; average = 12.2). 71 legal entities maintain a care should be taken that well-known visualiza- continuous flow over all 15 quarters to between 1 and tion techniques are chosen and the user interface 24 media. Media have between 1 and 1,577 incoming is self-explaining. flows (median = 1; average = 9.4). These flows connect R5 Interactive exploration: Some users will approach them to between 1 and 285 distinct legal entities (me- the MT visualization trying to verify an existing dian = 1; average = 3.2). 68 media maintain a contin- hypothesis but we expect that a majority of us- uous flow over all 15 quarters from between 1 and 18 age session will consist of undirected exploration legal entities. in search for patterns of interest. For this, interac- 67 Exploring Media Transparency With Multiple Views Table 2: Media matching the query string “standard” ters best [Mun14, p. 150]. A similar aggregation is ordered by the number of connected legal entities and visualized for sum of money transferred by legal back- showing the aggregated sum of transferred money. ground in Figure 2.B. Three entries are di↵erent forms of writing the same To visualize the distribution of a single quantitative website. The fifth entry contains the names of six sep- attribute a histogram can be used [Mun14, p. 306]. arate newspapers and stands as example of inconsis- Figure 2.C shows a histogram of the data attribute tent data collection when he MT dataset was started “money amount”. in 2012. Figure 2.D is a histogram of the data attribute medium #rel. summed flows “trend”. This attribute is derived from all amounts of money ei flowing from a legal entity to a medium Der Standard 189 18,905,741 e over time i. The trend T quantifies the relative di↵er- derstandard.at 64 2,768,875 e ence of money transferred between the first half of the www.derstandard.at 19 312,242 e quarters |Qm | and the second half. Der Standard KOMPAKT 2 44,745 e Standard Verlagsge- 1 3,099,082 e sellschaft m.b.H. j k |Q| Krone, Kurier, Presse, 1 90,874 e |Qm | = 2 (1) Salzburger Nachrichten, P|Q| P|Qm | 1 ei ei Standard, Kleine Zeitung T = P|Q| i=|Qm | i=1 (2) ei derstandard.at App 1 11,510 e i=1 ES Evening Standard 1 10,884 e Magazine The categorical data attributes legal entity and http://www.derstandard.at 1 9,938 e medium both have a large number of categories (see Subsection 3.2), which are too many to visualize them in a bar chart. Neither is it possible to aggregate the tive features are needed that are usable and help categories in a reasonable way. But the entries can users maintain overview. be sorted by another aggregated quantitative data at- tribute so that only the most relevant ones are dis- 4 Visualization Design played For example, it is possible to sort legal entities by the sum of money transferred from them to media. Based on these design requirements, we developed a vi- Figure 2.E shows the details for the first 10 sorted legal sualization design for the MT dataset (Figure 2). This entities as a table with 4 columns. The first column section describes the design and explains its underly- shows the name of the legal entity. The second column ing rationale. In Subsection 4.1 the individual diagram shows the sum of the transferred money to various me- views of the design are presented. How the user is dia over time. Additionally a sparkline sized bar chart able to interact with them is described in Subsection represents the distribution of the transferred money 4.2 and how the views are linked with each other is over time. This enables the user to see the trend over delineated in Subsection 4.3. time [Tuf06, AMST11]. The third column displays the number of relations, i.e. the count of media receiving 4.1 Attribute Visualization money from the legal entity. The forth column dis- The MT dataset contains 5 data attributes. The plays the “trend” as calculated by Equation 1. This columns of Table 1 display these data attributes. It is data table visualization enables the user to receive de- not possible to visualize every single data record of this tailed aggregated information for a few entities. Fig- table in the dashboard, therefore the records are ag- ure 2.F applies the same visual representation to the gregated and the aggregated information is displayed. first 10 sorted media. [Mun14, p. 305] The last visualization in Figure 2.G shows the flow For example Figure 2.A shows aggregated data of money from legal entities to media using a chord of money transferred over time. For this, the data diagram [KSB+ 09]. The aggregated amount of money attribute “money amount” is summed for all data is encoded with the length of an arc of a circle seg- records with the same value in the data attribute ment of the diagram. This allows the user to see from “time”. This reduces the data to only 2 data attributes which legal entity how much money is transferred to and only 1 data record per quarter. The sum is a quan- which medium. Because there are too many di↵erent titative attribute and the quarters can be handled as categories, placeholder segments are generated for le- ordinal attribute. A bar chart suites the task of look- gal entities and media, which contain all not displayed ing up and comparing the values of the di↵erent quar- entries and aggregate the money for all of them. 68 Exploring Media Transparency With Multiple Views A B C D H H I I E G F Figure 2: The visualization design for the media transparency (MT) dataset is comprised of seven views: (A) bar chart of aggregated money by time, (B) second bar chart for money by legal category, (C) histogram of money by flow, (D) histogram of increasing/decreasing trend, (E) table of 10 legal entities with total money, sparkline of money, number of connected media, and trend, (F) second table of top 10 media, and (G) chord diagram of flows. Both tables can be (H) searched and (I) sorted. 4.2 Interaction Components for every visualization. By clicking onto a data table row the entity is selected. The histograms The interaction with the diagrams is essential for the (Figure 2.C&D) do not support a simple click op- user to explore the data and to verify or refute an eration, but a click and drag operation to select initial hypothesis (R5). a one dimensional range of the attribute in the In Figure 2 the data is visualized without any ma- histogram [Mun14, ch. 11.4]. nipulations by the user. The first three diagrams (Fig- Highlight Elements To visualize which visual ele- ure 2.A,B,C&G) give the user an overview of the un- ments are selected, the color saturation of the vi- derlying data. To analyze the data further the user is sual element is increased and for the filtered ele- able to manipulate the view of the data. ments the saturation is decreased. Figure 3 shows Details on Demand The visualization design en- this di↵erence in saturation in contrast to the not ables the user to receive details of a visual en- selected visual elements in Figure 2.A&B. The col- coding of an aggregation of a data attribute of a oring of the small-multiple bar charts in the data chart. The visual encoding of a number, for ex- table is also linked with the highlighting of the ample the height of a bar in a bar chart or the time bar chart. The used colors are selected using length of an arc in a flow visualization, supports the ColorBrewer2 tool, which is based on evalu- the user to compare the encoding with the same ation of “385 unique colour schemes [...] across encoded data attributes. To receive exact num- di↵erent computer platforms and monitors, [...] bers the user is able to hover over each visual en- for possible colour-blind confusions, as well as in coded element and receive in place information printed formats.” [HB03] with a tool-tip [Dix09]. Sort To explore detail information for the trend over Select Elements To receive even more detailed in- time, money, and the number of relations from one formation about the highlighted visual element, entity to another, the user is able to sort the data the user is able to select it. The data is then fil- table along the data attribute of her/his interest tered by the selection and all other visualizations (see Figure 2.I). are updated with the newly filtered data. This in- Search To support users’ who want to analyze the teraction method is implemented in the visualiza- data for a specific entity, full-text search is inte- tion design as simple left-mouse-click and works 69 Exploring Media Transparency With Multiple Views grated. In our visual design this is implemented 6 Usage Scenario as a simple form text fields for the legal entities This sections presents a usage scenario to understand and media (see Figure 2.H). how the visualization of the MT dataset enables users Combine and Remove Like already mentioned in to obtain a deeper insight into the data. The steps Section 3.1 the data quality might not be optimal. of the scenario can be followed in a video located at As modifying the underlying data cannot be ex- https://vimeo.com/188278798. pected by the target user group, interactive visual In most cases a user that is interested in a data editing should be possible (R3). In our prototype, set has some a-priori knowledge and a hypothesis that users may remove entries and combine multiple she/he wants to verify or falsify. In this scenario the entries into a single entry. With a click onto the user is interested in which legal entities spend money labels above a data table the selected rows of that on online advertisement with Google. table are combined or removed. Entering “google” into the full text search, the list returns 57 entries (e.g.: google, google.at, www.google.at, ...) due to data quality issues. By 4.3 Coordinating Multiple Views interactively manipulating the data the user is able to The designed interface connects the di↵erent visual- obtain a deeper insight. For example by combining izations and widgets and organizes them. The views the 57 categorical entries of the data attribute media are arranged on fixed positions, but the user is able to one entry named “Google”. The flow visualization to filter the data [EB11, Rob07]. Because all visual- is now easier to read because the number of visual el- izations of the media transparency database use the ements was reduced. same data set it is possible to link the selection be- The user is able to filter data which she/he is not tween all views and thus use each view for dynamic interested in to obtain new information. For example query [AS94, ST98]. Additionally the color of the vi- by selecting only the entries of universities. This re- sual elements indicate which aggregation is used. This sults in 3 legal entities, which the user is now able to helps the user to see the connection between the vi- compare in more detail (see Figure 4). sualizations and it enables the user to understand the connection of a data attribute in one to the distribu- 7 Conclusions tion of a data attribute in another diagram [Mun14, This paper presented a visualization design to explore ch. 12]. the MT dataset, a large open government data asset reporting on the flows of money from government to media. We implemented the design as a web-based 5 Implementation prototype, made it publicly available, and showcased it on science communications events like Lange Nacht der The visualization design has been implemented as a Forschung. Based upon these experiences and informal web-based software using JavaScript with the libraries feedback we received, we can now reflect how well the D3.js [BOH11], Crossfilter [cro], and dc.js Dimensional visualization design addresses its design requirements Charting [dc]. and provide directions for future research: The implementation is available from https: R1 Data scalability: The various views of aggre- //github.com/VALIDproject/mtdb2 as free and gated attributes are useful to provide a big-picture open source software under a BSD-2-clause license overview of the dataset. Subsequently, the inter- and can be tested at http://medientransparenz. action concept of linked selection, sorting tables, validproject.at/dashboard/. For iterative refine- and showing the first results works to learn about ment an informal usability test with two subjects was the details. Some users criticized the chord dia- conducted. gram as being too cluttered and hard to read. A Sankey diagram is being considered as alternative. In future work, two additional proposals from the preceding problem characterization study by Niederer et al. [NRA+ 16] can be adopted: The large number of legal entities and media could be automatically clustered into hierarchical groups using text or network analytics. Alternatively, supplementary data could be loaded to provide Figure 3: The elements of the visualizations adapt additional properties such as geographic area for color saturation upon selection changes. legal entities and/or media. These properties 70 Exploring Media Transparency With Multiple Views Figure 4: Snapshot of the visualization after following the steps of the usage scenario. would subsequently be used for filtering and ag- should be provided so that these features are not gregation. visible by default. R2 Development over time: Both the bar chart view R4 Ease of use: The visualization design is built us- showing aggregated money flow over time, and ing simple visual representation techniques that the sparkline sized bar chart for each legal en- are well known to the general public. Still, the tity/medium work well to show distribution, ab- multiple views in composition were described as normalities and other temporal patterns for the slightly overwhelming at first impression. In ad- currently selected respectively visible items. The dition, novice users were not aware of direct ma- derived attribute “trend” was added to allow nipulation so they did not expect they could filter overview and direct manipulation of one concrete the data e.g. by clicking on a bar. temporal pattern. While being a powerful fea- R5 Interactive exploration: As demonstrated in the ture, it is hard to grasp for novice users of the MT usage scenario, the visualization design allows free dataset visualization. Further design experiments exploration of the MT dataset. While doing so, are necessary to provide user-friendly exploration users can maintain overview of system state, i.e. of temporal dynamic flows in bipartite networks. which selections are active and also reset selec- R3 Data wrangling: The interactions to combine legal tions. entities and/or media o↵ers some benefits. The As further support for exploration, the data jour- views are less cluttered by di↵erent entries for nalists interviewed by Niederer et al. [NRA+ 16] related institutions. In some cases data wran- suggested documentation of the research path in gling can eliminate a perceived false patterns such order to provide analytic provenance [NCE+ 11]. as abrupt end of flow to one medium that is in Thus, our design study yielded not only a possible fact continued to a medium of a slightly di↵erent visualization design but also a range of directions for name. future work on exploring flows in dynamic bipartite Further work on data wrangling is indicated: On networks. the one hand, we found the current functionality too limiting in several exploration sessions and de- Acknowledgements sired more flexibility such as hierarchical groups and/or multi-group assignment like tags. On the This work was supported by the Austrian Ministry other hand, the two functions ‘combine’ and ‘re- for Transport, Innovation and Technology (BMVIT) move’ introduced more confusion for first time under the ICT of the future program via the VALiD users. Possibly, a dedicated data wrangling mode project (no. 845598) and by the Austrian Science Fund (FWF) via the KAVA-Time project (no. P25489). 71 Exploring Media Transparency With Multiple Views References [dc] dc.js – dimensional charting Javascript library. https://dc-js.github.io/dc. [AMST11] Wolfgang Aigner, Silvia Miksch, Hei- js/. Accessed: 2015-11-16. drun Schumann, and Christian Tomin- ski. Visualization of Time-Oriented Data. [Dix09] Alan Dix. Human-Computer Interaction. Springer, London, 2011. Springer, 2009. [AS94] Christopher Ahlberg and Ben Shneider- [EB11] Micheline Elias and Anastasia Bezerianos. man. Visual information seeking: Tight Exploration views: understanding dash- coupling of dynamic query filters with board creation and customization for vi- starfield displays. In Proceedings of the sualization novices. In Proc. 13th IFIP SIGCHI conference on Human factors in TC 13 Int. Conf. Human-Computer In- computing systems: celebrating interde- teraction, INTERACT, pages 274–291. pendence, pages 313–317. ACM, 1994. Springer, 2011. [Aus15] Julian Ausserhofer. “Die Methode liegt [GBD09] Martin Greilich, Michael Burch, and im Code”: Routinen und digitale Meth- Stephan Diehl. Visualizing the evolu- oden im Datenjournalismus. In Axel tion of compound digraphs with TimeArc- Maireder, Julian Ausserhofer, Christina Trees. Computer Graphics Forum, Schumann, and Monika Taddicken, edi- 28(3):975–982, 2009. tors, Digitale Methoden in der Kommu- [Ham] Markus Hametner. Inserate: 40,6 Mil- nikationswissenschaft, Digital Communi- lionen im ersten Quartal. http:// cation Research, pages 87–111. Berlin, derstandard.at/2000017464403/. Web- 2015. Standard: 2015-06-15. [BBDW16] Fabian Beck, Michael Burch, Stephan [HB03] Mark Harrower and Cynthia A Brewer. Diehl, and Daniel Weiskopf. A taxon- ColorBrewer.org: An online tool for se- omy and survey of dynamic graph visual- lecting colour schemes for maps. The Car- ization. Computer Graphics Forum, pub- tographic Journal, 40(1):27–37, 2003. lished online before print: 25 January, 2016. doi: 10.1111/cgf.12791. [HBW14] M. Hlawatsch, M. Burch, and D. Weiskopf. Visual adjacency lists [BKH+ 16] Benjamin Bach, Natalie Kerracher, for dynamic graphs. IEEE Trans. Vi- Kyle Wm. Hall, Sheelagh Carpen- sualization and Computer Graphics, dale, Jessie Kennedy, and Nathalie 20(11):1590–1603, 2014. Henry Riche. Telling stories about dynamic networks with graph comics. [HSS15] Ste↵en Hadlak, Heidrun Schumann, and In Proc. CHI Conf. Human Factors in Hans-Jörg Schulz. A survey of multi- Computing Systems, pages 3670–3682. faceted graph visualization. In Rita ACM, 2016. Borgo, F. Ganovelli, and Ivan Viola, edi- tors, Proc. Eurographics Conf. Visualiza- [BOH11] M. Bostock, V. Ogievetsky, and Jef- tion – State of The Art Report, EuroVis frey Heer. D3: Data-driven documents. STAR, pages 1–20. Eurographics, 2015. IEEE Trans. Visualization and Computer [KSB+ 09] Martin Krzywinski, Jacqueline Schein, Graphics, 17(12):2301–2309, 2011. İnanç Birol, Joseph Connors, Randy Gas- coyne, Doug Horsman, Steven J. Jones, [CMS99] Stuart K. Card, Jock D. Mackinlay, and and Marco A. Marra. Circos: An informa- Ben Shneiderman, editors. Readings in tion aesthetic for comparative genomics. Information Visualization: Using Vision Genome Research, 19(9):1639–1645, 2009. to Think. Morgan Kaufmann, San Fran- cisco, 1999. [Lan] Fabian Lang. Medientransparenz - die zweite. http://www.paroli-magazin. [cro] Crossfilter – fast multidimensional fil- at/555/. Paroli-Magazin: 2013-03-18. tering for coordinated views. http:// square.github.io/crossfilter/. Ac- [Lor10] Mirko Lorenz. Status and outlook for cessed: 2015-11-16. data-driven journalism. In Data-driven 72 Exploring Media Transparency With Multiple Views journalism: what is there to learn?, pages [SBSV] Peter Salhofer, Amir Basyouni, Mer- 8–17. European Journalism Centre, 2010. cedes Stibler, and Stephan Vrecer. Me- dientransparenz Austria. http://www. [Med15] Medienkooperations- und -förderungs- medien-transparenz.at/. Accessed: Transparenzgesetz, MedKF-TG. 2016-09-09. Austrian Legal Information System, 2015. https://www.ris.bka.gv.at/ [SMM12] Michael Sedlmair, Miriah Meyer, and Dokumente/Erv/ERV_2011_1_125/ERV_ Tamara Munzner. Design study method- 2011_1_125.html. Accessed: 2016-08-31. ology: Reflections from the trenches and the stacks. IEEE Trans. Visualization [Mun14] Tamara Munzner. Visualization Analysis and Computer Graphics, 18(12):2431– and Design. CRC Press, 2014. 2440, 2012. [NAR15] Christina Niederer, Wolfgang Aigner, and [ST98] Robert Spence and Lisa Tweedie. The Alexander Rind. Survey on visualizing Attribute Explorer: Information synthesis dynamic, weighted, and directed graphs via exploration. Interacting with Comput- in the context of data-driven journalism. ers, 11(2):137–146, 1998. In Hans-Jörg Schulz, Bodo Urban, and Uwe Freiherr von Lukas, editors, Proc. [Tuf06] Edward Rolfe Tufte. Beautiful Evidence. Int. Summer School on Visual Computing, Graphics Press, Cheshire, CT, 2006. pages 49–58. Frauenhofer Verlag, 2015. [vdEvW14] Stef van den Elzen and Jarke J. van Wijk. [NCE+ 11] Chris North, Remco Chang, Alex Endert, Multivariate network exploration and pre- Wenwen Dou, Richard May, Bill Pike, and sentation: From detail to overview via se- Glenn Fink. Analytic provenance: pro- lections and aggregations. IEEE Trans. cess+interaction+insight. In Proc. 2011 Visualization and Computer Graphics, Ann. Conf. Ext. Abstracts Human Factors 20(12):2310–2319, 2014. in Computing Systems, CHI EA, pages 33–36. ACM, 2011. [vLKS+ 11] Tatiana von Landesberger, Arjan Kui- jper, Tobias Schreck, Jörn Kohlhammer, [NRA+ 16] Christina Niederer, Alexander Rind, Jarke J. van Wijk, Jean-Daniel Fekete, Wolfgang Aigner, Julian Ausserhofer, and Dieter W. Fellner. Visual analysis of Robert Gutounig, and Michael Sedl- large graphs: State-of-the-art and future maier. Visual exploration of me- research challenges. Computer Graphics dia transparency for data journalists: Forum, 30(6):1719–1749, 2011. Problem characterization and abstrac- tion. In Proc. 10th Forschungsforum der [WPZ+ 16] Yanhong Wu, Naveen Pitipornvivat, Jian österreichischen Fachhochschulen. FH des Zhao, Sixiao Yang, Guowei Huang, and BFI Wien, 2016. Huamin Qu. egoSlider: Visual anal- ysis of egocentric network evolution. [Rob07] Jonathan C. Roberts. State of the art: IEEE Trans. Visualization and Computer Coordinated & multiple views in ex- Graphics, 22(1):260–269, 2016. ploratory visualization. In Proc. Conf. Coordinated and Multiple Views in Ex- [ZGC+ 16] Jian Zhao, Michael Glueck, Fanny Cheva- ploratory Visualization, CMV, pages 61– lier, Yanhong Wu, and Azam Khan. Ego- 71, 2007. centric analysis of dynamic networks with EgoLines. In Proc. CHI Conf. Human [RTR] Austrian regulatory authority for broad- Factors in Computing Systems, pages casting and telecommunications. Bekan- 5003–5014. ACM, 2016. ntgegebene Daten. https://www.rtr. at/de/m/veroeffentl_medkftg_daten. Accessed: 2016-08-31. [RTR16] RTR GmbH. Katalog Medientransparenz. O↵ene Daten Österreich, 2016. https: //www.data.gv.at/katalog/dataset/ 58e02823-2bd2-4db7-9e2f-72a9ea7c7ffd. Accessed: 2016-08-31. 73