This tab is all about representing your results graphically. You should be able to generate publication-quality figures without needing to export the data from the tool.
Most options here are self-explanatory. Pick the results you want to plot, play with some options, and hit
Plot. This can be a great way to identify salient linguistic features in your dataset. Once you’ve plotted something, a navigation pane appears under the figure, allowing scrolling, zooming, saving and so forth.
Below are explanations of the various options.
Data to plot
Here, you select the interrogation or edited result you want to plot, as well its branch.
If you like, you can plot only a single entry or subcorpus from an interrogation or edited result.
Results to show
Plot only the top
n entries from the results of interest. If you don’t want these, go back to the
Edit tab and remove/sort the results some more. You can type in
all to get every result—but be warned, this often won’t look so good!
Kind of chart
You can choose between line, area, pie, bar or heat charts.
corpkit tries to guess useful axis labels. When it fails, you can enter your own.
When making pie charts, you can explode one or more results—that is, cut them slightly out of the pie for emphasis. Enter the name of the entry you want to explode, or use a list:
Or use a pre-defined wordlist:
You can plot using a logarithmic scale for either axis.
This option adds counts from previous subcorpora to later subcorpora, to give a cumulative distribution.
Temporarily flip the x and y axis. This can be performed when editing, too.
Black and white
Render the figure in black and white. For line charts, try to create unique-looking line markers.
You could also use a greyscale colour scheme with a line plot, in order to produce a black and white figure without the complex line types
Use TeX option will mean that the plotter will try to use
LaTeX to typeset the text in the chart. If you have a TeX distribution, but this option isn’t working for you, you may need to use the command-line version of
This option reverses the order in which legend entries occur. It can be useful when making horizontal bar (‘barh’) charts.
Subplots creates a separate figure for each entry in your results. You might have to resize the figure a little when using this option to give it sensible proportions.
Show or hide the grid in the figure’s background.
For bar charts, stack entries on top of each other, rather than beside each other.
If you’re making a pie chart, and the percentages don’t sum to 100%, this option will allow the plotting of pie ‘slices’, rather than entire pies.
When making stacked bar or area plots, you can use this option to make every entry sum to 100%.
Choose the colourmap for plotted entries.
Aesthetic changes to the plot: background colour, line size, etc. If using a seaborn style, colours cannot be manually selected.
Choose where in the figure the legend should sit, or let it float around with
Choose dimensions for the figure. Equal dimensions is probably a good idea for pie charts.
If you’re using the
Subplots option, this allows you to specify the grid shape for the canvas as a whole.
Try to display aggregates or total percentages on the plot, in the legend, or both.
Interacting with your figure
Once you’ve generated a figure, a toolbar appears beneath it, with options for panning and zooming. The figure is smart enough to reposition the legend depending on the position of the data currently being shown.
You can use this toolbar to save the image to disk. Your
project/images directory is a good place for them.
If you’re more confortable generating visualisations in another tool, you can head to
Manage project to export a CSV version of any interrogations or edited results. This can be imported into Excel, for example.
Viewing other images
You can use the
Next buttons to bring up any images that have been saved in the
Through the process of interrogating, editing and visualising, you may find something that sticks out: maybe something is very frequent in a subcorpus, or maybe something is conspicuously absent. The
Concordance tab allows you to resuse your
Interrogate queries to look at language in co(n)text.