CrossRE: Cross-Domain Relation Extraction
Cross-domain relation extraction across 6 domains (news, politics, science, music, literature, AI). Annotators identify entities and label 17 relation types between entity pairs, enabling study of domain transfer in relation extraction.
Configuration Fileconfig.yaml
# CrossRE: Cross-Domain Relation Extraction
# Based on Bassignana & Plank, Findings EMNLP 2022
# Paper: https://aclanthology.org/2022.findings-emnlp.263/
# Dataset: https://github.com/mainlp/CrossRE
#
# CrossRE provides relation extraction annotations across 6 diverse domains:
# news, politics, natural science, music, literature, and AI.
# This enables studying how relation extraction models transfer across domains.
#
# Entity Types:
# - Person, Organization, Location, Event, Date, Number, Other
#
# 17 Relation Types:
# - physical: Physical proximity or containment
# - personal-social: Personal or social relationships
# - part-whole: Component or membership
# - org-affiliation: Organizational affiliation
# - gen-affiliation: General affiliation
# - usage: One entity uses another
# - topic: Entity is about a topic
# - compare: Comparison between entities
# - origin: Entity originates from another
# - named: Entity is named after another
# - role: Entity has a role
# - temporal: Temporal relationship
# - win-defeat: Competitive outcome
# - located-in: Spatial containment
# - cause-effect: Causal relationship
# - member: Membership in a group
# - no-relation: No meaningful relation
#
# Annotation Guidelines:
# 1. Read the sentence and note the domain
# 2. Identify all named entities and classify them
# 3. For each entity pair, determine the relation type
# 4. Optionally add notes about ambiguous cases
annotation_task_name: "CrossRE: Cross-Domain 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: Identify entities
- annotation_type: span
name: entities
description: "Highlight all entities in the text"
labels:
- "Person"
- "Organization"
- "Location"
- "Event"
- "Date"
- "Number"
- "Other"
label_colors:
"Person": "#3b82f6"
"Organization": "#22c55e"
"Location": "#ef4444"
"Event": "#f59e0b"
"Date": "#8b5cf6"
"Number": "#06b6d4"
"Other": "#6b7280"
keyboard_shortcuts:
"Person": "1"
"Organization": "2"
"Location": "3"
"Event": "4"
"Date": "5"
"Number": "6"
"Other": "7"
tooltips:
"Person": "Names of people, including fictional characters"
"Organization": "Companies, institutions, bands, teams, political parties"
"Location": "Countries, cities, geographic regions, addresses"
"Event": "Named events, wars, festivals, elections, discoveries"
"Date": "Dates, years, time periods, centuries"
"Number": "Numerical values, quantities, measurements"
"Other": "Entities not fitting other categories (works of art, products, etc.)"
allow_overlapping: false
# Step 2: Link entities with relation types
- annotation_type: span_link
name: relations
description: "Draw relations between entity pairs"
labels:
- "physical"
- "personal-social"
- "part-whole"
- "org-affiliation"
- "gen-affiliation"
- "usage"
- "topic"
- "compare"
- "origin"
- "named"
- "role"
- "temporal"
- "win-defeat"
- "located-in"
- "cause-effect"
- "member"
- "no-relation"
tooltips:
"physical": "Physical proximity, containment, or spatial relationship"
"personal-social": "Personal or social relationship between people"
"part-whole": "Entity A is a part or component of entity B"
"org-affiliation": "Person or entity is affiliated with an organization"
"gen-affiliation": "General affiliation not covered by org-affiliation"
"usage": "Entity A uses or employs entity B"
"topic": "Entity A is about or concerns entity B"
"compare": "Entity A is compared with entity B"
"origin": "Entity A originates from or is derived from entity B"
"named": "Entity A is named after entity B"
"role": "Entity A has a role or function related to entity B"
"temporal": "Temporal relationship between entities"
"win-defeat": "Competitive outcome between entities"
"located-in": "Entity A is located within entity B"
"cause-effect": "Entity A causes or leads to entity B"
"member": "Entity A is a member of entity B"
"no-relation": "No meaningful relation between the entities"
# Step 3: Optional annotator notes
- annotation_type: text
name: notes
description: "Optional notes about ambiguous cases or difficult decisions"
html_layout: |
<div style="margin-bottom: 10px; padding: 8px; background: #f0f4f8; border-radius: 4px;">
<strong>Domain:</strong> {{domain}}
</div>
<div style="font-size: 16px; line-height: 1.6;">
{{text}}
</div>
allow_all_users: true
instances_per_annotator: 50
annotation_per_instance: 2
allow_skip: true
skip_reason_required: false
Sample Datasample-data.json
[
{
"id": "crossre_001",
"text": "Barack Obama served as the 44th President of the United States from 2009 to 2017, having previously represented Illinois in the U.S. Senate.",
"domain": "politics"
},
{
"id": "crossre_002",
"text": "The Beatles, formed in Liverpool in 1960, became the best-selling music act of all time, with estimated sales of over 600 million units worldwide.",
"domain": "music"
}
]
// ... and 8 more itemsGet This Design
Clone or download from the repository
Quick start:
git clone https://github.com/davidjurgens/potato-showcase.git cd potato-showcase/text/relation-extraction/crossre-cross-domain-relations potato start config.yaml
Details
Annotation Types
Domain
Use Cases
Tags
Found an issue or want to improve this design?
Open an IssueRelated Designs
MultiTACRED: Multilingual TAC Relation Extraction
Multilingual version of the TACRED relation extraction dataset in 12 languages. Annotators identify subject/object entities and classify relations from 41 TAC relation types, with additional translation quality assessment.
REDFM: Filtered and Multilingual Relation Extraction
Multilingual relation extraction across 20+ languages derived from Wikidata. Annotators verify entity spans and label relations from a curated set of 400+ Wikidata relation types, enabling large-scale multilingual relation extraction research.
Structured Fact Verification (FEVEROUS)
Verify claims against Wikipedia pages containing both unstructured text and structured data (tables, lists, infoboxes). Based on FEVEROUS (Aly et al., NeurIPS 2021). Annotators assess verdict, identify evidence across different modalities, evaluate evidence sufficiency, and provide justification.