Institut für Theoretische Informatik, Algorithmik

On the Readability of Boundary Labeling

The raw data of the user study on boundary labeling can be downloaded at data.csv.zip. Please refer to [1] to get a description of the study setup and a detailed discussion of the results.

The raw data is provided in CSV format. Every row in the dataset describes one solved task, i.e. one click made by a participant.

These are the fields of the CSV data:

  • generator: The generator / type of this boundary labeling. Possible values are „po“, „do“, „sl“ and „opo“.
  • x and y: The x and y coordinates at which the user has clicked. These are probably useless without having the original image. Look at on_target for more useful information. These values may be floats because of scaling.
  • time: The absolute time in miliseconds it took the user to click after the image was revealed to him.
  • difficulty: The difficulty of the presented image. May be „easy“, „med“ or „hard“. Please see the paper for what this means.
  • session: A unique ID for the session that this click is part of. Every participant has contributed exactly one session, so all clicks with the same session are from the same user.
  • sites: The number of sites that the task solved had. May be „30“ or „15“.
  • time_normalized: The click time normalized by the median click time of the corresponding session, i.e. the median click time of the user that clicked.
  • variance: The variance of the site cluster in this task. The value „0“ indicates a uniform distribution of sites, while „3000“ stands for a dense cluster of sites and „10000“ for a sparse cluster.
  • img: ID of the instance that was solved with this click
  • task: Whether the user was asked to assign a label to a given site („stol“) or a site to a given label („ltos“).
  • on_target: Whether we considered that click „correct“ or not. In the stol-case, a click was considered correct if its y coordinate was closer to the center of the correct label than to any other label. In the ltos-case, a click was considered correct if it was closer to the correct site than to any other site and also within a certain maximum distance from the correct site.

In case of questions please contact Benjamin Niedermann.

[1] L. Barth, A. Gemsa, B. Niedermann, M. Nöllenburg: On the Readability of Boundary Labeling In: Proceedings of the 23rd International Symposium on Graph Drawing (GD'15), Lecture Notes in Computer Science. Springer, 2015.