Chemical-Disease Relation Extraction (BC5CDR)

# Chemical-Disease Relation Extraction (BC5CDR)
# Based on BioCreative V CDR Task
# Guidelines: https://biocreative.bioinformatics.udel.edu/
#
# This task identifies chemical-induced disease (CID) relations from
# biomedical abstracts. A CID relation exists when a chemical is
# reported to cause or induce a disease.
#
# Entity Definitions:
# - Chemical: Drugs, small molecules, elements, chemical compounds
# - Disease: Diseases, disorders, symptoms, adverse effects
#
# Relation Types:
# - CID (Chemical-Induced Disease): The chemical causes or induces the disease
# - No Relation: Entities are mentioned but no causal relationship
#
# Annotation Guidelines:
# 1. First identify all chemical and disease mentions
# 2. For each chemical-disease pair, determine if a CID relation exists
# 3. CID requires evidence of causation (not just co-occurrence)
# 4. Include: adverse effects, side effects, toxicity
# 5. Exclude: treatments (chemical treats disease), preventions
# 6. Hypothetical or negated relations should not be annotated
#
# Evidence Patterns for CID:
# - "X causes Y", "X-induced Y", "X leads to Y"
# - "Side effects of X include Y"
# - "Y was observed in patients taking X"

annotation_task_name: "Chemical-Disease Relation Extraction"
task_dir: "."

data_files:
  - sample-data.json
item_properties:
  id_key: "id"
  text_key: "text"

output_annotation_dir: "annotation_output/"
output_annotation_format: "json"

annotation_schemes:
  # Step 1: Mark chemical entities
  - annotation_type: span
    name: chemicals
    description: "Highlight all CHEMICAL mentions (drugs, compounds, elements)"
    labels:
      - "Chemical"
    label_colors:
      "Chemical": "#3b82f6"
    tooltips:
      "Chemical": "Drugs, medications, chemical compounds, elements, or small molecules"
    allow_overlapping: false

  # Step 2: Mark disease entities
  - annotation_type: span
    name: diseases
    description: "Highlight all DISEASE mentions (diseases, disorders, symptoms, adverse effects)"
    labels:
      - "Disease"
    label_colors:
      "Disease": "#ef4444"
    tooltips:
      "Disease": "Diseases, disorders, medical conditions, symptoms, or adverse effects"
    allow_overlapping: false

  # Step 3: Identify CID relations
  - annotation_type: radio
    name: cid_relation
    description: "Is there a Chemical-Induced Disease (CID) relation in this text?"
    labels:
      - "Yes - CID relation present"
      - "No - No causal relation"
      - "Uncertain"
    tooltips:
      "Yes - CID relation present": "The text states or implies that a chemical causes/induces a disease"
      "No - No causal relation": "Chemicals and diseases are mentioned but no causation is stated"
      "Uncertain": "The relationship is ambiguous or hypothetical"

  # Step 4: Relation evidence
  - annotation_type: span
    name: evidence
    description: "Highlight the text that indicates the CID relationship (if present)"
    labels:
      - "CID Evidence"
    label_colors:
      "CID Evidence": "#22c55e"
    tooltips:
      "CID Evidence": "Words or phrases that indicate the chemical causes the disease"
    allow_overlapping: true

allow_all_users: true
instances_per_annotator: 50
annotation_per_instance: 2
allow_skip: true
skip_reason_required: false