Extraction Viewer enhancements

Safety One 2026.1.01 introduces enhancements to the Extraction Viewer (previously referred to as the annotation tool), including a redesigned interface, expanded attribute support, and improved usability for reviewing and validating AI-extracted data from unstructured text during intake processing.

Note:

AI-powered automatic extraction from source documents is for evaluation purposes only. Please review the output thoroughly before use.

You can access the redesigned Extraction Viewer in the Intake Form interface by selecting Extraction Viewer from the Case Files drop down menu in the document pane. You can use the extraction viewer inline in the document pane for a streamlined view or expand and detach the panel to see details like confidence scores, extraction relationships, and full attribute context.

• Separate extraction viewers may be listed in the Case Files drop down menu depending on the source type: Extraction Viewer (Attachment) and Extraction Viewer (Email).
• An expanded set of attributes is supported, so you can review and validate additional data points extracted from unstructured text (such as narrative and email body).
• Original formatting (such as spaces and line breaks) from unstructured text are retained in the Extraction Viewer.
• If a corresponding value in the Intake Form is modified or deleted, the related annotation is automatically removed from the extraction tool.

  Note:

  You cannot perform manual corrections or deletions of annotations directly in the Extraction Viewer.
• The Extraction Viewer is independent of the Intake Form Narrative field, so changes you make to the Narrative field do not affect the data displayed in the viewer.
• The Highlight All Extractions toggle in the Intake Form document pane highlights every data point extracted from the structured document, helping you quickly verify all information captured by the extraction engine within the original source.
• The system now supports the extraction of data from source documents and E2B files for all fields supported in the Intake Form.
• The drugs and events auto extracted using the Entity Extraction model (unstructured) or Advanced Data Extraction (structured) are coded using appropriate dictionaries.