5.9 Network
5.9.1 Options
Network tool has the following options.
Layout: by default, a layout is automatically selected, but you can manually select the one you prefer
Directional: if this is ON, an arrow indicates the directionality of edges; you can specify the Arrow Size
Remove - Loop Edges: when a pair of the same item (word) is on the list, this will be represented by a loop on the graph because the same item only appears once on the graph; with this option ON, loops will be removed from the graph
Remove - Multiple Edges: when Directional option is OFF and two pairs of the same items appear twice in different orders (i.e. 1 2 and 2 1), two edges are drawn between them; with this option ON, only one edge will be drawn between them
Remove - Ind. Pairs: with this option ON, pairs of items that are not connected to other nodes (or independent) will be removed from the graph
Cut off: with this option ON, edges and associated nodes with the specified criterion will be removed from the graph; options are Edge Freq, Betweenness, Edge Freq (Q), and Betweenness (Q); the first two are to be specified with an absolute value and the last two are to be specified with ration (0 to 1.0); if 0.5 is specified, values below 50 percentile will be eliminated
Node Size: by default you can specify the uniform size for all the nodes; with the box unchecked, you can specify the information to be used (only Degree is available for now) and how to transform the value;√(/π)or√√(/π)means a square root of the Node Frequency value divided by π (3.14...) and a square root of it, and Prop. means the size is proportional to the largest node (largest = 1); you can further adjust by multiplying, dividing, or raising by power with a specified value
Node Shape: you can specify the shape of nodes; if one of rectangle options is selected, specify the height
Edge Width: by default you can specify the uniform width for all the edges; with the box unchecked, you can specify the information to be used (FrequencyorBetweenness) and how to transform the value;√or√√means a square root of the selected value and a square root of it,, and Prop.means the width is proportional to the widest edge (widest = 1); you can further adjust by multiplying, dividing, or raising by power with a specified value
Node Label: whether or not to all labels to nodes
Node Label Offset: specify the offset value; a value above zero (0) offsets the label to upper right; a value below zero offsets the label to lower left
Edge Label: you can add labels on edges based onBetweenness or Raw Count of the edges;
specify the Label Size
Color: specify the color of Node, Label, and Edge; for Node, choices are the default color, a specified uniform color, specified colors on the color table, and colors assigned based on community membership; you can select how to group the nodes; for Label, choices are the default color, a specified uniform color, and specified colors on the color table; for Edge, choices are the default color, a specified uniform color, and gray scale based on Betweenness indexes Label Size: specify the size of labels
The following plots were drawn with the same sample data from FROWN corpus with different Layoutoptions. As noted above, each time you run the analysis, you will not have exactly the same plot in terms of positioning of nodes. The plots with some of the options look similar with this data
set, but might be quite different with other data sets.
Random Circle Sphere
Fruchterman-Reingold Kamada-Kawai Reingold-Tilford
LGL Graphopt
Handling Directionality, Loop, and Multiple (duplicate) Edged affects as follows:
No Options Directional with Loop Directional without Loop
Non-directional without Loop Non-directional without Multiple Non-directional without Both
Remove Ind. option eliminates the pairs of nodes each of which only has one connected node.
All nodes Independent node pairs eliminated
Cut offoption eliminates the edges below the set value. With Edge Freq (Q), the value should be specified between 0.0 and 1.0. 0.0means all the edges will be kept.1.0means the edge with the most frequency will be kept.
0.0 0.5 0.7 0.9
In the next two examples, the one on the left was drawn with fixed node size and edge width. The one on the right was drawn with theFixedoption ofNode SizeandEdge WidthOFF. TheArrow Size was also adjusted to match the edge width.
Fixed Node size and Edge width Variable Node size and Edge width
Shapeoptions are as follows. At this moment,pieoption does not do much. Drawing the content of pies require the values to represent proportions of somethings. The examples so far used circle option.
square rectangle
pie sphere
You can draw small nodes and offset the labels as in the following plot.
Edge Label can be either betweenness or frequency of the edge.
Betweenness Count (Frequency)
Coloring of Nodes, Labels, and Edges has few options each. The default coloring of nodes is Select with a single color ‘lightblue’. You can change the color on the pop-up button.
If you select specify, specify the colors for individual nodes on the coloring table. Click Color button on the panel and click Fill to fill the table with node candidates. The number of items to appear on the table depends on theLimityou set on the pane. So make sure you do not increase the Limit once the color assignment is done. The following assignments resulted in the plot on the right.
If you select Community, community membership is assigned to each node based on the selected method. As you can see in the image below, some methods are only available for non-directional network.
With the above options, the bi-gram list of FROWN corpus with independent node pairs eliminated, the plot will look like below (left). You can also change the Label colors in the same way (right).
For the Edge color, if you select Gray (Betweenness), the shades of gray represent the betweenness values. Higher values are represented by darker shades.
If you combine these options and color the nodes by POS information, a plot like below can be drawn. The assignment of colors was done on Label Coloringpanel (see Section 5.12.13). This in fact is a part of larger network, plotted on a larger area (14 x 14 inches) using a bi-gram list excluding ones with stop words. The source of the text is obvious if you have read them.
The lighter green nodes are person’s names and the orange nodes are verbs. The largest (most frequent) node is ‘said,’ which is connected to many person’s names. From the direction of arrows, it can be inferred that the frequent form is “~, said ‘a person’”, not “‘a person’ said ~.” Another large node is ‘harry,’ which is connected to many verbs in past tense. “Professor” is another word which is connected to the names of ‘professors’ at Hogwarts.