Text Preprocessing Examples

Examples

This topic describes part of the functionality of Genesys Content Analyzer.

The table "Examples of Preprocessing Filters" displays simple examples of text-preprocessing filters.

Examples of Preprocessing Filters

Pattern Type	Pattern Body	Input Text	Test Result
DELETE AFTER	finch	one two finch three four	one two
DELETE BEFORE	finch	one two finch three four	three four
	[Mm]essage_?[Ss]tart	x897 message_Start one two three	one two three
DELETE ALL IF FIND	finch	one two finch three four	footnote a.
		one two three four	one two three four
	internal\d\d	one two three internal36 four
DELETE ALL IF NOT FIND	finch	one two finch three four	one two finch three four
	finch	one two three four
DELETE PATTERN	f.*ch\s	one two finch three four	one two three four
		one two fach three four	one two three four

a. If you test this filter, the resulting window contains the message TEXT HAS BEEN DELETED. In actual use of DELETE ALL IF FIND or DELETE ALL IF NOT FIND, the entire text object is deleted from the training object.

Tables "Preprocessing Filters Example" and "Results of Testing the Example" present a more complex example using all five filter types.

The table, "Preprocessing Filters Example" lists the filters used in the example.

Preprocessing Filters Example

Filter Number	Pattern Type	Pattern Body
1	DELETE BEFORE	MessageStart
2	DELETE AFTER	IDnumber=
3	DELETE ALL IF FIND	internal\d\d
4	DELETE ALL IF NOT FIND	nihil_obstat
5	DELETE PATTERN	company

The table "Results of Testing the Example" shows an example of input text and the results of applying the filters from "Preprocessing Filters Example" to it.

Results of Testing the Example

Input Text	Test Result
x88_2 MessageStart nihil_obstat: Hello, companyyes, good-bye.IDnumber=7989	nihil_obstat: Hello, yes, good-bye.

The results in the table "Results of Testing the Example" come about as follows:

1. Filter 1 deletes the text "x88_2 MessageStart."
2. Filter 2 deletes the text "IDnumber=7989."
   
3. Filter 3 does nothing (it finds a match for nihil_obstat).
   
4. Filter 4 does nothing (it fails to find a match for internal/d/d).
   
5. Filter 5 deletes "company."