Skip to content
SmiKar Software

Get Label Suggestions from Forage

5 min read · Last updated · Page version 9

Where content classification is enabled on a deployment.

Purpose: let Forage suggest labels for the kinds of record your sites hold. The AI assistant, which runs inside your own deployment, does the looking; you decide.

  1. Open Records, then Suggested Labels, and press Suggest labels.

  2. Tick the sites to look through, or Every included site (one run can cover up to 1,200,000 files). Only sites Forage has read are listed, with how many files it has read in each. If your choice holds more files than that, Forage still starts: it takes each site's share of the limit in proportion to the files read there and says how many of your choice it looked through. You can also choose the kinds of file to look at, such as documents and spreadsheets only.

  3. Press Start. Leave Email me when they're ready filled in to be emailed when the suggestions are ready to review, with how many labels were suggested, how many files they cover and a link to this page; it holds the address you used last, or the one you signed in with, and appears once emailed reports are set up. While a run goes, its panel shows who will be emailed, with Change to use another address or Don't email to send none. Reading slows down while Forage looks, and returns to normal speed by itself when the suggestions are ready.

  4. Follow the steps on the page: making room, getting ready, AI reading the groups, putting it together and AI checking the suggestions. The page shows the time the run has left. You can cancel at any time.

  5. When the suggestions are ready, each card shows a suggested label, how many files it covers, where they are, and the keywords a rule would look for. A line above the cards says how many of the files read fell into kinds of record with enough files alike to suggest a label for, how many sat in smaller groups or on their own, and how many had too little text to group. Forage has already tried the keywords against the files in the chosen sites: the card says how many files a rule on them would find, how many are in the group and how many outside it, and whether a sample of the outside files fit. Press Show the files to list them in the card with their folder, type, size, date and site, Open in Search to see the files the rule would reach, or Show example files to open some of the files behind the suggestion.

  6. For each suggestion, choose one:

    • Accept creates the label, with the name, category, severity, description and keywords as you edit them, and a draft rule that gives the label to files containing all of those keywords together;
    • Use existing label saves that draft rule for a label you already have;
    • Merge folds it into another suggestion for the same kind of record;
    • Reject sets it aside, and later suggestions leave it out; Put back for review on a rejected suggestion returns it to the list so you can accept, edit, merge or reject it again;
    • Edit changes its name, description, category, severity, the words a file must contain and the words that rule a file out, and shows how many files the words would find as you type.

    Run again repeats a run with the same sites and kinds of file, which you can change before starting.

  7. Open Content Rules, give the new draft rule a test run, and switch it on when you are happy with what it finds. Until then it labels nothing.

The Suggested Labels page: tiles for the suggestions to review, the files covered out of those read in the chosen sites, the suggestions accepted and the kinds of record found; then a card for each suggestion with its name, what it covers, its severity and category, how many files it covers and in how many groups, how many files a rule on its keywords would find and how many of a sample fit, the keywords a rule would look for, the sites they are in, a link to example files, and buttons to accept it, use it for an existing label, merge it or reject it

Good to know: Forage groups files by what their text contains, never by their names, and a suggestion comes from a group of files that are alike, so a file unlike the others in its sites may get no suggestion. A suggestion's category and severity are a starting point: check them before you accept. Every decision is recorded on the Log page.

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.

More in Squirrel

See all pages →