To train a new model, there are the following two options:
You can schedule training that uses an existing scheduled job as a template. This is a convenient way to change the time that an existing job is scheduled to run. To do this:
On the Training Schedule tab, right-click the existing job that you want to use as a template.
From the context menu, select New Training Job (Use this Job as Template).
You can create a new training job from scratch. To do this, on the Tools menu of the Training or Training Schedule tab, select Schedule New Model Training.
Both options bring up the Model Training Schedule dialog box. With option 1, the fields of the dialog box are populated with values from the existing job. With option 2, the fields are populated with default values.
The Model Training Schedule dialog box has two tabs, Model Options and Text Preprocessing.
Enter a name for the model that will result from the scheduled training.
Select a training object from the drop-down list.
Select the treatment of the Subject field of e-mails:
Ignore
Add to the
text
Add with
double weight
Select training quality. Draft is lowest quality, 6 is highest. The higher the quality, the longer training takes.
Select either no cross-validation or cross-validation that splits the data into three, five, or ten sets.
Cross-validation is a way of producing accuracy ratings of the model at the same time that the model is being produced. Splitting the data into more sets produces more accurate ratings, but it also makes training take more time. For explanation see "Cross-Validation" in the "Genesys Knowledge Management: Content Analyzer" chapter of the Multimedia 7.6 User's Guide.
Enter a start time or select a unit (day, month, hour, minute) and change the unit using the up and down arrows.
Enter a number for minimum samples in category; that is, the minimum number of text objects that a category must have in order to be included in training.
Enter a keyword threshold; that is, the minimum number of text objects in which a key word must occur for that keyword to be considered in training.
Select whether to use all categories for training or terminal categories only (a terminal category is one that contains no subcategories).
Leave Training Data Quality set at Regular unless you know that the training object contains many wrongly categorized text objects.
Click Add Filter. The New Filter dialog box appears.
Select filter type from the following:
DELETE AFTER—delete all text after and including the matching text.
DELETE BEFORE—delete all text before and including the matching text.
DELETE ALL IF FIND—delete all text from the e-mail that includes the matching text.
DELETE ALL IF NOT FIND—delete all text if the e-mail does not include the matching text.
DELETE PATTERN—delete only any text matching the pattern.
Enter a filter body. This can be either a literal text string or a regular expression. See Constructing Screening Rules and the "Genesys Knowledge Management: Basics" chapter of the Multimedia 7.6 User's Guide for information on regular expressions.
Return to the Text Preprocessing tab.
Select a filter in the list and click Delete Filter to delete or Edit Filter to edit.
Click Move Up and Move Down to impose order on multiple filters.
In the Test Filter area, enter some sample text and click Test to see the result of applying the filter.
See also "Schedule Training" in the "Genesys Knowledge Management: Content Analyzer" chapter of the Multimedia 7.6 User's Guide.