Skip to main content

Annotation Tool

The Annotation Tool is a proposed tool to allow experts to semantically annotate free text in ePIs with standard terminology labels, serving as manual preprocessing or generating training data for NLP models.

Purpose​

Enables human experts to:

Features​

Text Selection​

  • Highlight text spans in ePI narrative
  • Select existing HTML elements (<p>, <h1>, etc.)
  • Create new semantic boundaries
  • Nested annotation support

Find and apply codes from:

  • SNOMED-CT: Clinical concepts
  • ICPC-2: Primary care classification
  • LOINC: Laboratory observations
  • ATC: Medication classification

Features:

  • Free-text search
  • Hierarchy browser
  • Recent/favorite codes
  • Synonym matching

Annotation Management​

Create Annotations​

  1. Select text span
  2. Search terminology
  3. Choose concept code
  4. Define elementClass name
  5. Preview in context
  6. Save annotation

Edit Annotations​

  • Modify concept code
  • Adjust text boundaries
  • Change elementClass
  • Add notes/rationale

Validate Annotations​

  • Check for overlaps
  • Verify code appropriateness
  • Review completeness
  • Export validation report

Output Formats​

Generate p(ePI) compatible output:

  • HtmlElementLink extensions
  • HTML with class attributes
  • FHIR Bundle JSON
  • Training dataset format

Collaboration Features​

Support multiple annotators:

  • Assign sections to experts
  • Track annotation progress
  • Compare inter-annotator agreement
  • Resolve conflicts
  • Consensus building

Training Data Export​

For NLP services:

  • Export as IOB/BIO format
  • JSON-LD for entity recognition
  • Custom ML framework formats
  • Split train/test/validation sets

User Roles​

Medical Expert​

  • Annotate clinical concepts
  • Validate terminology selection
  • Review automated annotations

Content Curator​

  • Manage annotation projects
  • Assign tasks
  • Monitor progress

ML Engineer​

  • Export training data
  • Validate dataset quality
  • Integrate with NLP pipelines

Integration​

Connects to:

Quality Assurance​

Features:

  • Inter-annotator agreement metrics (Cohen's Kappa)
  • Annotation guidelines enforcement
  • Consistency checks
  • Expert review workflow