Fine-Grained Propaganda Detection
Span-level annotation of propaganda techniques in news articles based on SemEval 2020 Task 11 (Da San Martino et al., EMNLP 2019). Identifies 14 techniques including Loaded Language, Name Calling, Appeal to Fear, and more across rhetorical dimensions (ethos, logos, pathos).
Archivo de configuraciónconfig.yaml
# Fine-Grained Propaganda Detection
# Based on Da San Martino et al., EMNLP 2019 and SemEval 2020 Task 11
# Paper: https://aclanthology.org/D19-1565/
# Task: https://propaganda.qcri.org/semeval2020-task11/
#
# 14 propaganda techniques organized by rhetorical dimension:
#
# PATHOS (Appeal to Emotions):
# - Loaded Language: emotionally charged words
# - Name Calling/Labeling: assigning negative labels
# - Exaggeration/Minimisation: amplifying or downplaying
# - Appeal to Fear/Prejudice: creating worry using biases
# - Flag-Waving: exploiting national/group pride
# - Slogans: brief striking phrases
# - Doubt: undermining credibility
#
# LOGOS (Appeal to Logic):
# - Causal Oversimplification: single cause for complex issues
# - Black-and-White Fallacy: false binary choices
# - Thought-terminating Cliches: phrases that stop critical thinking
# - Straw Man/Whataboutism/Red Herring: misrepresentation or deflection
#
# ETHOS (Appeal to Authority):
# - Appeal to Authority: citing experts without evidence
# - Bandwagon/Reductio ad Hitlerum: everyone does it / Hitler did it
annotation_task_name: "Fine-Grained Propaganda Technique Detection"
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:
- annotation_type: span
name: propaganda_technique
description: "Highlight text spans and label them with the propaganda technique used"
labels:
# Pathos - Appeal to Emotions
- Loaded Language
- Name Calling/Labeling
- Exaggeration/Minimisation
- Appeal to Fear/Prejudice
- Flag-Waving
- Slogans
- Doubt
# Logos - Appeal to Logic
- Causal Oversimplification
- Black-and-White Fallacy
- Thought-terminating Cliches
- Whataboutism/Straw Man/Red Herring
# Ethos - Appeal to Authority
- Appeal to Authority
- Bandwagon
- Repetition
label_colors:
"Loaded Language": "#ef4444"
"Name Calling/Labeling": "#f97316"
"Exaggeration/Minimisation": "#eab308"
"Appeal to Fear/Prejudice": "#84cc16"
"Flag-Waving": "#22c55e"
"Slogans": "#14b8a6"
"Doubt": "#06b6d4"
"Causal Oversimplification": "#0ea5e9"
"Black-and-White Fallacy": "#3b82f6"
"Thought-terminating Cliches": "#6366f1"
"Whataboutism/Straw Man/Red Herring": "#8b5cf6"
"Appeal to Authority": "#a855f7"
"Bandwagon": "#d946ef"
"Repetition": "#ec4899"
tooltips:
"Loaded Language": "Words with strong emotional connotations to influence the audience (e.g., 'brutal regime', 'freedom fighters')"
"Name Calling/Labeling": "Assigning labels that the audience despises or fears (e.g., 'radical', 'extremist')"
"Exaggeration/Minimisation": "Making something bigger or smaller than reality (e.g., 'the greatest disaster ever')"
"Appeal to Fear/Prejudice": "Building support by creating anxiety, often leveraging existing biases"
"Flag-Waving": "Exploiting patriotism or group pride to promote a position (e.g., 'true Americans believe...')"
"Slogans": "Brief, striking phrases that combine labeling with emotional appeal (e.g., 'Make X Great Again')"
"Doubt": "Questioning someone's credibility or character to undermine their argument"
"Causal Oversimplification": "Attributing complex problems to a single cause (e.g., 'immigration is why...')"
"Black-and-White Fallacy": "Presenting only two choices when more exist (e.g., 'you're either with us or against us')"
"Thought-terminating Cliches": "Phrases that discourage critical thinking (e.g., 'it is what it is')"
"Whataboutism/Straw Man/Red Herring": "Deflecting by changing the subject or misrepresenting the opponent's position"
"Appeal to Authority": "Claiming truth based on expert endorsement without evidence"
"Bandwagon": "Suggesting everyone supports something to pressure agreement (e.g., 'millions of people agree')"
"Repetition": "Repeating a message until it becomes accepted through familiarity"
allow_overlapping: true
allow_all_users: true
instances_per_annotator: 50
annotation_per_instance: 3
allow_skip: true
skip_reason_required: false
Datos de ejemplosample-data.json
[
{
"id": "prop_001",
"text": "The radical left wants to destroy everything our great nation stands for. True patriots must stand up and fight back before it's too late!"
},
{
"id": "prop_002",
"text": "Scientists agree that climate change is real. The data clearly shows rising temperatures over the past century."
}
]
// ... and 10 more itemsObtener este diseño
Clone or download from the repository
Inicio rápido:
git clone https://github.com/davidjurgens/potato-showcase.git cd potato-showcase/text/fact-verification/propaganda-techniques potato start config.yaml
Detalles
Tipos de anotación
Dominio
Casos de uso
Etiquetas
¿Encontró un problema o desea mejorar este diseño?
Abrir un issueDiseños relacionados
Clickbait Spoiling
Classification and extraction of spoilers for clickbait posts, including spoiler type identification and span-level spoiler detection. Based on SemEval-2023 Task 5 (Hagen et al.).
Detecting Persuasion Techniques in News
Identification of propaganda and persuasion techniques in news articles through both multi-label classification and span-level detection. Based on SemEval-2023 Task 3 (Piskorski et al.).
Patronizing and Condescending Language Detection
Detection and span-level identification of patronizing and condescending language directed toward vulnerable communities in news media. Based on SemEval-2022 Task 4 (Perez-Almendros et al.).