The information about the tool is described in the paper
ScanGraph: A novel scanpath comparison method using graph cliques visualization
(Dolezalova and Popelka, 2016)
You can download the article by clicking to the PDF icon.
Dolezalova, J. & Popelka, S. (2016). ScanGraph: A novel scanpath comparison method using graph cliques visualization. Journal of Eye Movement Research, 9(4), 13p. doi:10.16910/jemr.9.4.5
Since the publication of the paper, we have implemented several changes of the tool:
07/2016 - Heuristic algorithm is not needed anymore. The cliques are found with the use of Bron-Kerbosch algorithm.
08/2016 - Damerau algorithm was added
09/2016 - Multiple file upload in zip support
10/2016 - Add mean, occurrance method for multiple file upload
Input data (OGAMA)
ScanGraph application is designed to work with data exported from open-source application OGAMA (www.ogama.net). The user can calculate similarity according to a regular grid or defined Areas of Interest.
In Scanpath module, OGAMA contains a tool called “Levenshtein Distance Calculation”, which is capable of computation of Levenshtein distances between participants’ trajectories.
For the use of ScanGraph application, only Subject names (1), Scanpath strings (2) and affiliation to subject groups (3) will be used.
Text file exported from OGAMA can be directly used in ScanGraph
Note: Uploaded data are stored anonymously at our server and will not be used for any other purposes.
String-edit distance method
The similarity can be calculated using Levenshtein, Damerau-Levenshtein, or Needleman-Wunsch algorithm. All algorithms were modified to better work with sequences with different lengths.
distance algorithm counts a number of needed changes (deletion, substitution, insertion) from one sequence to another.
algorithm counts a number of needed changes (deletion, substitution, insertion, transposition) from one sequence to another.
algorithm seeks the number of identical elements of the pair of sequences.
Calculation of Needleman-Wunsch and Damerau-Levenshtein algorithms take more time, the user has to wait for the result (in some cases more than 30 seconds).
Using this button, the user can choose between original and collapsed fixation sequences. In collapsed sequences, there are no successive AOIs in the sequence.
Multifile method selection
Choose of method of multifile computation:
ScanGraph will compute the mean of modified matrices for all used stimuli. OCCURRANCE:
Vertices are connected if they are connected in given % (degree of occurrance) of graphs for each stimuli.
This button returns a graph with 5 % of possible edges and a corresponding value of parameter
. This graph is user-friendly and according to our experiences has a very high interpretive value about the similarities.
This option is recommended for users with no experiences with ScanGraph
User defined graph
Degree of occurrance <0,1>
Define degree of occurrance
Use parameter p <0,1>
takes value from the interval <0,1> and represents the degree of similarity.
The higher value of
, the higher similarity of the given sequences.
Use the density of a graph [%]
Percent of edges takes value from the interval <0,100>. The complete graph contains 100 % of edges.
According to our experiences, the value 5 % has a very high interpretative value about the similarities.
© 2015 - 2020
& Ondřej Štrubl, Palacký University Olomouc, Faculty of Science,
Department of Geoinformatics