CHAPTER 1 - Text Analysis (part VI)

Equivalent classes and Relations graphs

For the graph display, you can choose between the following modes: [Area], [Star], [Distribution] or [Episodes]. Use the context menu that appears when clicking on an object of the graph with the right button of the mouse.

The first two graphs (area and star) show the Relations between the Equivalent classes. They are oriented: the classes on the left of the central class are its predecessors, those on the right its successors.

The first graph displays the classes in the form of Areas:

Each Equivalent class appears as a sphere, whose surface is proportional to the number of words it contains.

The distance between the central class and the other classes is proportional to the number of Relations connecting them: in other words, when two classes are close together, they share many Relations, and when they are far from one another, they share few Relations.

If we take up the metaphor about the planetary system, we can say that there is a central planet (Equivalent class), around which planets (Equivalent classes having Relations with the central class) of various sizes (according to the number of word occurrences they contain) revolve more or less closely (are more or less frequently used together).

Notes:

1 - The overlapping of two spheres has no particular significance.

2 - Word categories cannot be displayed in the area mode.

The Star graph displays the Relations between the Equivalent classes, or between a Word category and an Equivalent class:

In the above example, the "liberty" semantic class (central point) is preceded by the "right", "human rights", etc. classes. On the right are many successors to "liberty": "declaration", "opinion", "difference", etc.

The figures shown on the graph give the number of Relations (co-occurrence frequency) existing between the various Equivalent classes.

You can follow the Relations shown on the graph by clicking directly on the classes you wish to study. This very powerful function enables you to move through a text while viewing its micro-worlds and analyzing the connections between its various actors.

When you use the [Relations] command of the Result frame, the graph reveals two central classes:

In this example, we can see that there is only one Relation leading from the "dictatorship" class to the "human rights" class, whereas there are two Relations leading from the "liberty" class to the "declaration" and "opinion" classes.

The distribution graph enables you to display a histogram showing the distribution of an Equivalent class, of a Relation (i.e. between two Equivalent classes), or of a Word category:

In the above example, we can see that the "Education" Equivalent class appears rather at the end (left side) than at the beginning (right side) of the text.

This graph is obtained by splitting the text into several sectors containing the same number of words, and by calculating the occurrence frequency of the selected Equivalent class within each sector. The histogram bars present each sector in chronological order, from the left (beginning of the text) to the right (end of the text). The dotted line indicates the average size of the histogram bars.

When clicking on a given bar, the display will automatically position itself on the propositions appearing from this point of the text onwards. To return to the standard display, make a second click on the same bar. When pointing at a given part of the histogram, the message line displays the number of words contained by the selected sector, as well as its place (beginning and end of the sector, calculated in number of words from the beginning of text).

The number of histogram bars and the sum of words contained in each bar are automatically determined by the software according to the total of words in the text, and to the size of the main window.

When the distribution graph concerns a Relation, the histogram gives the accumulated occurrence frequencies of the classes contained in the Relation.

---