Navigating Contextual Complexity with Graph Visualization
The sheer size of the knowledge graphs we interact with is one form of complexity . However , aside from the problem of data volume , each of the significant variables or facets may need to be addressed as a unique abstraction of the data , or may need to be combined , as with change over time combined with geospatial awareness . As we understand geospatial data in the context of maps or renderings that are spatially proportionate , it is also the case that many other domains provide either physical or conceptual maps or metaphors as well , and representing these conceptual maps is likely necessary to effectively present data to users in those domains .
2 NAVIGATION AND GRAPH STRUCTURES
Many of us have been conditioned as users of technology to navigate with trees , which afford movement up and down through hierarchical structures , but to move sideways , to another branch , you may be taking a big move that fundamentally changes your context and point of reference . Other types of graphs , on the other hand , facilitate movement in any direction afforded by the graph , which can make maintaining that point of reference even more challenging if the right contextual clues are not present . von Landesberger et . al . [ 7 ] present a very thorough overview of graph-based display strategies and interaction methods . They further describe navigation strategies as falling into one of three categories : top-down , bottom-up , or middle-out .
Top-down navigation starts with the whole graph and affords navigation to drill into the data . Identification of a node of interest may involve some form of visual analysis . Top-down may be the preferred starting point for data science tasks , or system monitoring tasks where the whole , or some abstraction of the whole system informs the point of interest .
Figure 2 – 1 is an example of top-down navigation in a network display that uses edge color to indicate bandwidth stress . Tabular display of connection details supports selection and isolation of one more nodes within the network for deeper analysis . The overview display supports orientation from the zoomed-in portion of the graph to the wider network . Legends and filters are displayed on the left-hand side .
Journal of Innovation 73