Conveying Uncertainty in Archived War Diaries with GeoBlobs

GeoBlobs visualize uncertain and ambiguous spatio-temporal data derived from handwritten War Diaries from the First World War (WWI), documenting the story of the British Army and its units on the Western Front (Grayson, 2016). We propose a design space that shows the possible variations of GeoBlobs and how these variations can communicate (un)certainty visually.

Handwritten military diary (left), GeoBlob design space (middle), GeoBlob example map with pictograph overlay (right).

Data, Uncertainties, and Challenges
Since 2014,
The National Archives (UK) has undertaken the digitization of more than a million analog diary pages, using an interactive crowd-sourcing platform (https://www.operationwardiary.org/). During this process, essential information about military units, including labels for casualties, unit strength, weather, everyday army life, soldiers names and grades, military activities at the front and non-military activities behind the line, have been tagged and annotated. However, the transcription process introduced uncertainty on many levels due to missing records or pages, misspellings, illegible passages or lost diaries.

Some uncertainties derive from the particular type of data that has been gathered, the circumstances of its creation, and its post-processing. These may be the result of human errors: the British soldiers who wrote the diaries made misspellings, faithfully transcribed, while the crowd-workers also introduced new typos, leading to further ambiguity regarding names of people and locations. Or they may be intrinsic to the historical record, as where spatio-temporal uncertainty arises when several places are mentioned on one single day in the diary. Close reading is then required to determine in which order the places were visited by the troops or whether a place was mentioned as a troop location or for other reasons.
These types of uncertainty may lead to ambiguous geographic coordinates during the geo-referencing process (assigning a numeric geographic location to a place name). As part of the collaboration between
The National Archives and
City, University of London, we developed GeoBlobs, an abstract representation of spatio-temporal data, which visualizes uncertain spatio-temporal data derived from the handwritten military diaries.

GeobBlob Design Space
GeoBlobs offer one possible solution to the problems that arise from uncertainty in the data. GeoBlobs are an abstract representation of moving entities on a map with uncertain positions. Instead of showing a unit at a given point in time, GeoBlobs convey an unordered estimation of the possible locations over a temporal window using enclosed shapes, or blobs (Collins et al., 2009). To this end, we apply heuristics to weigh each location within the temporal window. At the representation stage, it is possible to specify dynamically which (weighted) sets of location are considered to form the GeoBlobs.
Different form and style parameters can influence the visual appearance of GeoBlobs and their semantics. The use of multiple GeoBlobs allows us to compare units and show different probabilities. Animation or overlays communicate additional context like military activities (e.g. fighting, raiding, fixing trenches) or non-military activities (e.g. playing football, resting, washing).
For example, three degrees of spatial certainty are represented by three superimposed GeoBlobs. Using the proposed design space, one can vary the visual attributes to convey uncertainty. For example, the shapes and their outlines can have different opacity or different levels of sketchiness or blurriness (Wood et al., 2012). While a crisp shape edge conveys certainty, a sketchy (hand-drawn) edge expresses a higher degree of uncertainty. Rather than offering a direct mapping between numerical probabilities and visual representations, the design space shows the possible variations of GeoBlobs and how these variations can communicate (un)certainty visually.

Application Examples
The GeoBlob project aims to reveal stories of the soldiers’ day-to-day life behind the lines, which will lead towards a narrative visualization (Riche et al., 2018) for communicating the life behind the trenches that cannot be found in our history books. A first approach we present is a comic-like sequence showing the day-to-day life of 3 battalions of the 3rd Division during WWI, which is communicated through text, GeoBlobs, and Isotype-like activity overlays (Liem et al., 2018a). We further developed an interactive, web-based prototype allowing to track and explore the activities and whereabouts of units over the course of the war (Liem et al., 2018b). A non-spatial map layer shows the distribution of a troop’s activities behind the line and at the front within a given time range. In future work, we plan to enrich the prototype by adding more data e.g., about single soldiers mentioned in the diaries.
GeoBlobs also apply to areas where the movement data is not uncertain but has a high density, or where the coverage area is of interest like sports data visualization (Perin et al., 2018) (e.g., visualization of motion, spatial coverage, pressure in team sports, and visual aggregation of players' spatio-temporal trajectories).

Bibliography
Collins, C., Penn, G. and Carpendale S. (2009). Bubble Sets: Revealing Set Relations with Isocontours over Existing Visualizations.
IEEE Transactions on Visualization and Computer Graphics, 16(6): 1009-1016.

Grayson, R. (2016). A Life in the Trenches? The Use of Operation War Diary and Crowdsourcing Methods to Provide an Understanding of the British Army’s Day-to-Day Life on the Western Front.
British Journal for Military History, 2(2): 160-185.

Liem, J., Goudarouli, E., Hirschorn, S., Wood, J. and Perin, C. (2018a). Conveying Uncertainty in Archived War Diaries with GeoBlobs.
IEEE VIS 2018 Electronic Conference Proceedings.

Liem, J., Goudarouli, E., Hirschorn, S., Wood, J. and Perin, C. (2018b). WWI00 {World War One Hundred}.
IEEE VIS 2018 Arts Program demonstrations.

Perin, C., Vuillemot, R., Stolper, C. D., Stasko, J. T., Wood, J. and Carpendale S. (2018). State of the Art of Sports Data Visualization.
Computer Graphics Forum, 37(3): 663-686.

Riche, N.H., Hurter, C., Diakopoulos, N. and Carpendale, S. (eds.) (2018).
Data-Driven Storytelling. Milton: Chapman and Hall/CRC.

Wood, J., Isenberg, P., Isenberg, T., Dykes, J., Boukhelifa N. and Slingsby A. (2012). Sketchy Rendering for Information Visualization.
IEEE Transactions on Visualization and Computer Graphics, 18(12): 2749-2758.