Kofax Getting Started with Ascent Xtrata Pro User Manual

Page 114

Advertising
background image

Classification

Ascent Xtrata Pro User's Guide

95

Statistics for each class are displayed when Statistics is selected from the toolbar. It is
also possible to save these results in text file format.

The Min. Confidence and Min. Distance sliders allow you to interactively modify
both thresholds after the result matrix has been calculated. Any changes you make
are immediately reflected in the matrix. This allows you to optimize the precision or
recall value by adjusting the confidence and distance threshold.

Note

After you have adjusted the thresholds to your liking, you might want to use

their values for the current project. To do this, the values determined in the result
matrix must be manually inserted as global thresholds for the Content Classifier in
the Project Settings dialog box. See Multipage Evaluation on page 63 for information
about the Project Settings dialog box.

The result matrix for a reference set can be calculated with or without hierarchical
rules. Without hierarchical rules, classification is faster, but hierarchical rules and
classification scripts are not applied during classification. This might lead to different
results as compared to standard classification.

For classification without hierarchical rules, the confidence and distance threshold
sliders are available in the result matrix for adjusting in order to optimize the
precision and recall. This classification mode is only available for classifying text
(*.txt) or image (*.tif) files. This means that you cannot classify XDoc files without
hierarchical rules.

The calculation of a result matrix with hierarchical rules uses the standard
classification mode, including classification script events. In this case, the threshold
sliders are not available because classification uses the thresholds from the project
settings. If you want to calculate the result matrix with different thresholds, you have
to modify the thresholds inside the project settings and then recalculate the result
matrix. This classification mode is available for all file types.

X

To calculate the result matrix on the reference set with hierarchical rules

1

From the main menu bar, select Tools | Calculate Result Matrix
for | Reference Set (hierarchical). A folder selection dialog box will display.

Advertising