Discover Research Tools for Textual Study
Knots is a visualization tool that helps to understand patterns of word relevance in one or more documents. Each term is represented as a twisted line – when the lines overlap it means a relevance or linkage within the terms.
Knots is an unusual visualization environment, one of the many text analysis tools offered as part of the Voyant toolset. All of the Voyant tools are web-based, and intended for use by both expert and non-expert users. You can either enter the URL of a web-based text, or upload one or more local documents for analysis in Knots.
When the tool initially loads it will by default display the most frequent words from the first document in a corpus. To enter new keywords, first remove existing terms using the “clear terms” button, and then enter new terms one at a time. Clicking on a selected term will toggle it on or off in the visualization. Unfortunately, the Knots tool cannot analyse terms across the whole corpus simultaneously, but evaluates only one document at a time. Use the “Documents” drop down menu to select a new document for analysis.
Each of the selected keywords is shown as a continuous line. Frequent terms will produced lines that are twisted and convoluted, whereas straighter lines indicate less frequent terms. Overlapping of lines indicates a linkage between the terms.
As you move your cursor over the knots, you will see that each is composed of a series of distinct line segments. Each of these segments represents one instance of the term. When clicked these segments will trigger a small window showing you the text excerpt in which that particular instance of the word occurs. Clicking a line segment will also open a new tab showing the selected term in the full Voyant text analysis interface.
Although we know that overlapping lines indicate some kind of relevance between the two terms, it is not possible to easily see exactly how the works are linked. It would be useful to have more information about the meaning of these overlaps. Are the words frequently found in close proximity? Do the words co-occur in a section or paragraph? The tool documentation provides little clue as to the type of relationship that might be represented by an overlap between two lines.
The “Tangles” slider controls how many twists there are within the visualization. When I moved the slider to all the way to the left to choose a value of 5 twists, the tool took longer to render the image than it had when the default value of 30 was selected. The effect of choosing a lower number of twists is to produce larger, looser knots that more easily expose the line segments that compose the knot, allowing for more granular exploration of a given term.
The following image shows the frequency of five different terms in a document using the lowest possible setting on the Tangles slider:
Here are the same terms from the same document rendered at the highest possible setting on the Tangles slider:
You can see that the representations are quite different. When the tangle setting is lower it is easier to distinguish each separate line segment in knots of high frequency words.
The Build Speed slider controls the speed at which the graphic is rendered. The rendering is fastest when the slider is pushed all the way to the left (lower number), and slowest when pushed all the way to the right (higher number). This is somewhat counterintuitive, as one expects the build to speed up as the slider is pushed to the right.
The “Starting Angle” slider adjusts the angle of the lines. It is useful for changing the orientation of the visualization for best fit within a browser window, and to create either a landscape or portrait version of the visualization for publishing in other sources.
Knots is a strange and interesting tool. It is most useful for comparing frequency between two or more separate terms in a given document. Knots provides some indication as to which of those terms are most closely linked to one another, although it is difficult to understand much about the nature of those relationships from within the Knots interface. Click through to the full Voyant interface for much more sophisticated analytics.
TAPoR v.2.5 | Copyright © 2014 TAPoR Team, University of Alberta.