AI Custom Fields/Tagging

This SOP explains how to create an AI-generated custom field for a technology taxonomy and use AI tagging to classify patents at scale. It also covers how to refine settings, use training data, filter by tags, and analyze results for reporting.

• Open a new folder containing the patent set you want to categorize.
  
  Go to the Tagging section.
  
  Select AI-generated custom field.
  
  Enter a descriptive technology-focused title, such as a product area, technology type, or innovation category.
  
  Use a clear prompt so the AI generates a relevant taxonomy.
• Review the AI-generated hierarchy once it appears.
  
  Confirm that parent and child categories match the intended technology structure.
  
  Edit category names if needed.
  
  Remove categories that are not relevant to your use case.
  
  Keep only the nodes you want to use for tagging.
• Decide whether the custom field should be Global or Workspace-specific.
  
  Use Global if tags should transfer across workspaces for the same patent.
  
  Use Workspace-specific if tags should remain only in the current workspace.
  
  Click Accept/Create to finalize the field.
  
  Confirm the new custom field appears in the spreadsheet.
• If needed, open Set field display in the view settings.
  
  Locate the new custom field in the list.
  
  Drag it to the preferred position, such as the top of the spreadsheet.
  
  Save or exit the display settings once the layout is correct.
• Go to the Tagging section again.
  
  Select AI tagging.
  
  Choose the custom field you want the AI to apply.
  
  Confirm you are working on the correct taxonomy before proceeding.
• Review the tag settings on the right-hand side.
  
  Confirm the patent reading scope.
  
  Select or deselect patent sections to include in the AI decision process.
  
  Choose a tag strategy:
  
  Single value if each patent should receive only one tag.
  
  Multivalue if a patent may belong to multiple nodes.
  
  Adjust the tag threshold based on your accuracy vs. coverage needs.
• Click Interpret Values to generate AI definitions for each node.
  
  Review the interpretation results carefully.
  
  Edit the free-text definitions if needed.
  
  Make sure each node definition is clear and aligned with your taxonomy goals.
• After reviewing the settings and definitions, start the tagging process.
  
  Allow the AI to analyze each patent against the custom field definitions.
  
  Confirm that tags are applied automatically to the best-matching technology category.
  
  Look for the AI indicator in the top-right corner of tagged records to verify completion.
• If you already have manually tagged patents, use them as training data.
  
  Confirm AI-generated tags where appropriate so they become training examples.
  
  Aim for at least 15 manually confirmed patents per node for better results.
  
  Return to AI Tagging and select the same custom field.
  
  Switch the model to Machine Learning when using training data.
  
  Start tagging again to allow untagged or unconfirmed patents to be tagged or retagged.
• Open the filter list and locate the custom field.
  
  Select a specific node, such as a technology subcategory, to narrow the document set.
  
  Apply the filter as a temporary rule for focused review.
  
  If the field is not visible, check the More Filters section.
• Go to the Analysis tab.
  
  Use data visualization tools to review patents by technology grouping.
  
  Analyze portfolio coverage, innovation trends, and category distribution.
  
  Use the results to support reporting and strategic decision-making.