Tamr Documentation

Filtering Records in Categorization Projects

Filter record categorizations by impact, assignment progress, response categorization, verification status, user, category, and source.

Filtering to High-Impact Records

  • High-impact records are those records in categorization projects from which the Tamr model learns the most.
  • After each update of categorization results, Tamr marks new records as being high-impact.
  • High-impact records comprise the top 1% of all ranked records, excluding verified records.
  • The default sort on the categorized records page is by impact.

To filter to high-impact records:

  1. Navigate to the categorized records page in a categorization project by selecting the option to the right of Categories. (The name of this page is project-specific and reflects what it is you are categorizing.)
  2. Select the filter icon on the top left-hand side.
  3. Choose High-Impact under the High-Impact Records header.

Filtering To Records that Need Verification

To filter to all records that have received input from reviewers but have not been verified:

  1. On the categorized records page, select the filter icon on the top left-hand side.
  2. Choose Needs Verification.
  3. Optionally, refine this filter further using option in the Categorizations section, based on whether records require further verification:
  • To filter to all records that can be verified to the Top expert suggestion without needing a manual review, turn off Users disagreed on categories.
  • To find records that need manual review before verification, turn on Users disagreed on categories and turn off Users suggested single category.

Filtering Records by Assignment Progress

To help you track assignment progress, Tamr includes the following filter options:

  • Not Assigned: Find records that are not assigned to any users.
  • Assigned: Find records assigned to at least one user.
  • Everyone responded: Find records with responses from all assigned users.
  • Some responded: Find records where at least one assigned user has not responded.
  • No one responded: Find records where no assigned users responded.

To filter records by assignment progress:

  1. On the categorized records page, select the filter icon on the top left-hand side.
  2. Scroll to the Assignments section.
  3. Choose one or more of Not Assigned, Assigned, Everyone responded, Some responded, and No one responded.

Filtering Records by User Response

To help you filter verified records by user feedback, Tamr includes the following filter options:

  • No user feedback: Find records with no user feedback.
  • User feedback: Find records where one or more users gave feedback on the record categorization.
  • Skipped: Find records where one or more users did not know how to respond.
  • Users suggested single category: Find records where all users agreed with the suggested category.
  • Users disagreed on categories: Find records where users suggested different categories, or where at least one user disagreed with a suggested category.

To filter records by user response:

  1. On the categorized records page, select the filter icon on the top left-hand side.
  2. Scroll to the Categorizations section.
  3. Choose one or more of No user feedback, User feedback, Skipped, Users suggested single category, and Users disagreed on categories.

Filtering Records by Verification Status

To help you track verification status on records in your dataset, Tamr includes the following filter options:

  • Not verified: All records suggested by Tamr or a reviewer, but not confirmed.
  • Verified: Verified record categorization. Tamr learns from verified categorizations only.
  • Agrees with Tamr: The verified categorization agrees with Tamr's categorization.
  • Disagrees with Tamr: The verified categorization disagrees with Tamr's categorization.
  • No Tamr suggestion: A categorization has been verified, but Tamr has not yet offered a categorization.

To filter records by verification status:

  1. On the categorized records page, select the filter icon on the top left-hand side.
  2. Scroll to the Verification section.
  3. Choose one or more of Verified, Not verified, Agrees with Tamr, Disagrees with Tamr, and No Tamr suggestion.

Filtering Records by Confidence

Confidence is a metric you can use to gauge how sure the Tamr model is in its categorizations for each record. Note that Confidence filters appear only after Tamr has predicted categorizations.

  1. On the categorized records page, select the filter icon on the top left-hand side.
  2. Scroll to the Confidence section.
  3. Chose one or more of High, Medium, Low, and Custom Range.
  4. If you select custom range, a slider appears to help you choose a specific range of confidence to run the filter on. When this filter is active, it works together with any other active confidence filters.

Filtering Records by Comments

  1. On the categorized records page, select the filter icon on the top left-hand side.
  2. Scroll to the Comments section.
  3. Choose Has Comments.

Filtering Records by an Assigned User

To filter records by the assigned team member:

  1. On the categorized records page, select the filter icon on the top left-hand side.
  2. Scroll to the Assigned to section (at the bottom) and choose Select users.
  3. Select a username and choose Update Filter.

Filtering by Category

Filtering by category includes both verified and unverified categories.

To filter records by category:

  1. On the categorized records page, select the filter icon on the top left-hand side.
  2. Scroll to the bottom and choose Add category under the Categories section.
  3. Select a category and choose Select.

Filtering Records by Source

To filter records by their source:

  1. On the categorized records page, select the filter icon on the top left-hand side.
  2. Scroll to the Source section (at the bottom) and choose Select sources.
  3. Select one or more input datasets, then choose Apply.

Updated a day ago


Filtering Records in Categorization Projects


Filter record categorizations by impact, assignment progress, response categorization, verification status, user, category, and source.

Suggested Edits are limited on API Reference Pages

You can only suggest edits to Markdown body content, but not to the API spec.