Using Custom Classifiers for Discovery

You can create custom classifiers to account for unique data samples present in your organization's data sources. This allows you to detect common data using a combination of patterns or dictionary values to classify them based on your supplied conditions or matching requirements. Users can also copy the detector template of a System Classifier to improve or customize the detector options available for those labels.

For more information on using our available System Classifiers in the V2 experience, see Using System Classifiers for Discovery.

To filter the Classifiers list screen

  1. On the Data Discovery menu, select Setup > Classifiers. The Classifiers list screen appears.

  2. Click the Add filter button to open a filter pop-up to choose whether to sort the list screen by Source or by Associated term.
    Source

    choose between System or Custom.

    Associated term

    choose from the available terms to pull all related classifiers (e.g., Phone Number to surface multiple phone number labels).

  3. After you select the filter criteria, the list updates automatically. To clear the selected filter criteria, click Reset selection.

To create a custom classifier

  1. On the Data Discovery menu, select Setup > Classifiers. The Classifiers screen appears.

  2. Click the Add classifier button. The Enter classifier details screen appears.

  3. Enter a unique name for the new classifier and an optional description. Classifier names can contain letters, numbers, spaces, and the following special characters: . ' ( ) - _
  4. Enable the Map new classifier to existing glossary term setting to map this classifier to an existing Glossary Term. Ignore if you wish to generate a new, identical term for mapping in the Glossary.


  5. Click the Next button. The Create classifier detectors screen appears.

To configure classifier detectors

To configure classifier rules

  1. Click the Add rule button.
  2. Configure the new rule, then click the Save button. The new classifier will be added to the main list in the Classifiers screen.

    Pattern-based rule



    Dictionary-based rule



    Field
    Description
    Rule name

    Enter a unique name for this new rule.

    Detector type

    Pattern: enter a RegEx pattern.

    Dictionary: enter all keywords (comma-separated) for data matching. Input is case sensitive unless Ignore case is enabled.

    Flag (applicable to pattern-based rule only)

    Choose from the following:

    • Global: Makes the expression search for all occurrences.

    • Ignore casing: Makes the expression search without case sensitivity

    • Multiline: Makes the boundary characters ^ and $ match the beginning and ending of every single line instead of the beginning and ending of the whole string.

    • Dot all: Makes the wild character '.' match newlines as well.

    • Unicode: Match with full Unicode.

    Set target

    Select a target option to evaluate the detection rule.

    • Finding

    • Context (requires before and after values)

    • Xpath

    • File extension

    • Object parent

    • File or column name

    • File name

    • Column name

    Choose your target conditions

    • Set likelihood: set target as Very likely, Likely, Possible, Unlikely, or Very unlikely if the rule matches, or No match if it does not match.

    • Adjust likelihood: adjust to More likely or Less likely if the rule matches or doesn't match.

    For more information, see Understanding Classification Likelihood Results.

To change a detector type's result

Important:

You can only modify the results choice of a detector of a custom classifier. Since system classifiers cannot be edited, you will need to copy the system classifier's detector type template when creating a custom classifier.

  1. On the Data Discovery menu, select Setup > Classifiers. The Classifiers screen appears.

  2. Select a custom classifier. The Classifier details screen appears.

  3. Go to the Detectors tab.
  4. Adjust the result preference in the Overall likelihood for the detector matching field. You can have the result be unique for each type.
  5. Click the Save button.