Write, Test and Switch On a Content Rule

Where content classification is enabled on a deployment.

Purpose: tell Forage what to look for, check what it would find, and switch it on.

1. Open Records, then Content Rules, and press New rule.
2. Give it a Name and choose the label it gives under Label to give. If there is no label yet, create one first.
3. Under Search words, write the words that pick out likely files, the way you would write a search, for example passport | "date of birth".
4. Add at least one detector under Detectors, and choose its type:
   • any word: any one of the words or phrases, matched as whole words;
   • all words: every one of the words or phrases must appear somewhere in the file;
   • pattern: a pattern anywhere in the text, with an optional Number check (payment cards, international bank account numbers, Australian business numbers, UK NHS numbers or Australian tax file numbers), so a value that fails the check is not reported.
5. To require words nearby, fill in Nearby words and Within this many characters. A match then counts only with one of those words close to it. Press Add detector for more.
6. Optionally press Add condition to add File conditions: file name, folder path, file type, size, modified date, read status, reason not read, where it is kept or site. Every condition must hold. A rule made only of conditions decides on those facts alone, so it can label files Forage could not read, such as every password-protected file.
7. Press Test run. It shows how many files contain the rule's search words, which is the most it could find, a sample checked against the full rule with what was found in context, and a projected range across the files read so far. Nothing is saved or labelled.
8. Press Save rule. The rule is saved inactive.
9. Select the rule to see its detectors and versions. On the version you want, press Activate, then Activate this version. Any other live version of the rule is retired at the same time, and Forage examines everything it has read again, in the background.
10. To change a rule, press Edit as new version, or New version from this on an older version, make your changes, press Save as new version, and activate it. The version you started from is not changed.
11. To switch a rule off, press Retire, then Retire this version. What it already found is kept.
12. A version that has never been active can be removed with Delete. A version that has been active is kept as history and can only be retired.
13. To check how often a rule is right, press Check accuracy on its card. Forage picks up to 25 of the files the active version labelled, at random. Open each one and press Right, Wrong or Not sure, then save the check. You can save before you have answered them all. The card then shows the share that was right, for example 88% right in a check of 25 files, and its tooltip shows the range the true figure is likely in, because 25 files is a sample. Files marked not sure are left out. Every check is in the Log, file by file, and belongs to the version it was made on, so check again after you activate a new version.
14. To tune a rule, select it and press Suggest words. Forage lists the words that are common in a random sample of up to 25 files the version labelled and uncommon in your other files, with how many of the sampled files contain each word and how common it is elsewhere. A word the rule already uses is marked and cannot be ticked. Tick the words you want and press Add to a new version: the rule editor opens with them added, and nothing is saved until you press Save as new version. Press Test run to see what the new version would find before you activate it.

Good to know: the test run is offered, not required, and can be run on any version from its row. A suggested-words list is a set of ideas from a small random sample, so it differs a little each time you ask, and a version that has not labelled any files yet has nothing to learn from. While Forage checks everything again after a change, the checked figure on Labels falls and then climbs again. A pattern that matches digits needs a checksum or context terms.

See what each page shows for the rest of the Records section, and product features for how each feature behaves.

Questions go to sales@smikar.com.