Analysis of Classes Page
The Analysis of Classes page displays the results of the analysis performed on your text classes in this experiment.
| Element | Description |
|---|---|
| Expected Accuracy |
The degree of accuracy to be expected when you classify a test text. For more details, see Expected Accuracy. |
| Top Features of Each Class |
The five most significant features of each text class. To see all examples of the feature within the class, click the feature name. Note: The Examples page displays the examples grouped according to the text class or test text in which they were found. Click a navigation link at the top of the Examples page to see the associated examples. For more details on features, see Features and Feature Sets. |
| View/Edit Features |
Edit your selection of feature sets and of individual features. For more details, see View and Edit Features. |
| View Outliers |
Opens the list of chapters the program did not correctly classify during testing of its classification model. For more details, see Outliers. |
What does the data tell us?
The data provided by class analysis can be helpful in a number of ways:
- If the expected accuracy is high, the data allows you to see which features contributed to that high figure. If it appears that there are many features that contribute to this determination, then this data can contribute to an argument that the two texts truly reflect different styles, and hence, possibly different authors. However, if it appears that the determination is made on the basis of a relatively small number of features, that argument becomes less compelling. For an example of this scenario, see Expected Accuracy.
- Identifying which features contributed to a high expected accuracy, can spur further avenues of exploration: what was the distribution of this feature within the text class? Are there perhaps other biblical texts where this feature occurs equally frequently, perhaps suggesting a connection between them?
- If the expected accuracy is low, this means that Tiberias has difficulty distinguishing between your text classes. You may consider: is this because one text class was composed as a close mimic of the other? Is this because, perhaps, the two text classes are really from a single source?
While Tiberias cannot provide definitive answers to these questions, its data can contribute to hypotheses about such authorial questions.