Connotation Frames of Power and Agency
Annotate verbs for implied power and agency of participants. Based on Sap et al., EMNLP 2017. Capture how actions implicitly convey power dynamics and agency levels.
text annotation
Configuration Fileconfig.yaml
# Connotation Frames of Power and Agency
# Based on Sap et al., EMNLP 2017
# Paper: https://aclanthology.org/D17-1247/
#
# Connotation frames capture how verbs implicitly project different
# levels of power and agency onto their participants.
#
# Key Concepts:
# - Agency: How much control/initiative does a participant have?
# - High agency: "She decided", "He led"
# - Low agency: "She was told", "He followed"
#
# - Power: Who has power over whom in the interaction?
# - Agent has power: "She commands him"
# - Theme has power: "She obeys him"
# - Equal: "They collaborate"
#
# Applications:
# - Detecting gender bias in media portrayal
# - Analyzing power dynamics in narratives
# - Understanding implicit framing in text
#
# Annotation Guidelines:
# 1. Focus on what the VERB implies, not the sentence context
# 2. Consider: Does this verb make the doer seem active/passive?
# 3. Consider: Does this verb give the doer power over others?
# 4. Think about typical usage, not edge cases
port: 8000
server_name: localhost
task_name: "Power and Agency Frames"
data_files:
- sample-data.json
id_key: id
text_key: sentence
output_file: annotations.json
annotation_schemes:
# Step 1: Agency of the subject
- annotation_type: radio
name: agency
description: "How much AGENCY (control, initiative) does the verb imply for the subject?"
labels:
- "High agency"
- "Moderate agency"
- "Low agency"
- "Neutral"
tooltips:
"High agency": "The subject is active, in control, making decisions (e.g., 'decides', 'leads', 'creates')"
"Moderate agency": "The subject has some control but within constraints"
"Low agency": "The subject is passive, reactive, or acted upon (e.g., 'follows', 'receives', 'is told')"
"Neutral": "The verb doesn't clearly convey agency level"
# Step 2: Power dynamics
- annotation_type: radio
name: power
description: "Who has POWER in the relationship implied by the verb?"
labels:
- "Subject has power"
- "Object has power"
- "Equal power"
- "No power dynamic"
tooltips:
"Subject has power": "The subject dominates or controls the object (e.g., 'commands', 'criticizes')"
"Object has power": "The object has power over the subject (e.g., 'obeys', 'serves')"
"Equal power": "Neither has power over the other (e.g., 'collaborates', 'discusses')"
"No power dynamic": "The verb doesn't imply a power relationship"
# Step 3: Sentiment toward subject
- annotation_type: radio
name: subject_sentiment
description: "Does the verb cast the subject in a positive or negative light?"
labels:
- "Positive"
- "Neutral"
- "Negative"
tooltips:
"Positive": "The verb portrays the subject favorably"
"Neutral": "The verb is neutral toward the subject"
"Negative": "The verb portrays the subject unfavorably"
# Step 4: Confidence
- annotation_type: likert
name: confidence
description: "How confident are you in your annotations?"
min_value: 1
max_value: 5
labels:
1: "Very uncertain"
2: "Somewhat uncertain"
3: "Moderately confident"
4: "Confident"
5: "Very confident"
allow_all_users: true
instances_per_annotator: 100
annotation_per_instance: 3
allow_skip: true
skip_reason_required: false
Sample Datasample-data.json
[
{
"id": "paf_001",
"verb": "commands",
"sentence": "The manager commands the team to finish the project."
},
{
"id": "paf_002",
"verb": "accepts",
"sentence": "She accepts the decision made by the committee."
}
]
// ... 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/power-agency-frames potato start config.yaml
Details
Annotation Types
Domain
Use Cases
Tags
Found an issue or want to improve this design?
Open an IssueRelated Designs
Argument Reasoning Comprehension (ARCT)
Identify implicit warrants in arguments. Based on Habernal et al., NAACL 2018 / SemEval 2018 Task 12. Given a claim and premise, choose the correct warrant that connects them.
Claim Perspectives (Perspectrum)
Annotate diverse perspectives on claims with stance and evidence. Based on Chen et al., NAACL 2019. Identify supporting and opposing perspectives for controversial claims.
Commonsense Inference (ATOMIC 2020)
Annotate commonsense inferences about events, mental states, and social interactions. Based on ATOMIC 2020 (Hwang et al., AAAI 2021). Generate if-then knowledge about causes, effects, intents, and reactions.