SemEval-2022 Task 5: Multimedia Misogyny (MAMI)
MAMI is the SemEval-2022 Task 5 benchmark for detecting misogynous memes from text and image content (Fersini et al.). This Potato config reproduces both subtasks: the binary misogyny label and the four-way misogyny type.
About this dataset
Multimedia Automatic Misogyny Identification (MAMI) was SemEval-2022 Task 5, organized by Elisabetta Fersini, Francesca Gasparini, Giulia Rizzi, Aurora Saibene, Berta Chulvi, Paolo Rosso, Alyssa Lees, and Jeffrey Sorensen. It was presented at SemEval-2022, the 16th International Workshop on Semantic Evaluation, held in Seattle in July 2022 and published by the Association for Computational Linguistics.
The dataset covers memes that pair an image with overlaid text. The shared task released 10,000 memes for training and 1,000 for the test set, each labeled for misogyny from both visual and textual cues.
Subtask A asks whether a meme is misogynous or not, a binary judgment. Subtask B applies only to misogynous memes and identifies the type among four overlapping categories: stereotype, shaming, objectification, and violence. A single meme can carry one or several of these labels, so Subtask B is multi-label. The data supports research on multimodal hate detection and content moderation.
The Potato config below reproduces both subtasks: a radio scheme for the binary misogyny label and a multiselect scheme for the four Subtask B types, with the meme text shown to the annotator.
- Venue
- SemEval-2022 Task 5 (Seattle, July 2022)
- Training memes
- 10,000
- Test memes
- 1,000
- Subtasks
- A: misogynous yes/no; B: type (multi-label)
- Subtask B categories
- stereotype, shaming, objectification, violence
- Participation
- 400+ participants, 65 teams in A, 41 in B, 13 countries
Configuration Fileconfig.yaml
This Potato config reproduces the annotation task. Save it as config.yaml and run potato start config.yaml to try it.
# MAMI - Multimedia Automatic Misogyny Identification
# Based on Fersini et al., SemEval 2022
# Paper: https://aclanthology.org/2022.semeval-1.74/
# Dataset: https://competitions.codalab.org/competitions/34175
#
# This task asks annotators to identify misogynistic content in memes
# by analyzing both the text overlay and the image description. If
# misogynistic, annotators classify the specific sub-types present.
#
# Binary Classification:
# - Misogynistic: The meme contains misogynistic content
# - Not Misogynistic: The meme does not contain misogynistic content
#
# Sub-type Labels (select all that apply):
# - Stereotype: Reinforces gender stereotypes
# - Shaming: Body-shaming or slut-shaming
# - Objectification: Treats women as objects
# - Violence: Promotes or trivializes violence against women
annotation_task_name: "MAMI - Multimedia Automatic Misogyny Identification"
task_dir: "."
data_files:
- sample-data.json
item_properties:
id_key: "id"
text_key: "text"
output_annotation_dir: "annotation_output/"
output_annotation_format: "json"
port: 8000
server_name: localhost
annotation_schemes:
- annotation_type: radio
name: misogyny_label
description: "Does this meme contain misogynistic content?"
labels:
- "Misogynistic"
- "Not Misogynistic"
keyboard_shortcuts:
"Misogynistic": "1"
"Not Misogynistic": "2"
tooltips:
"Misogynistic": "The meme contains content that is hateful, demeaning, or discriminatory toward women"
"Not Misogynistic": "The meme does not contain misogynistic content"
- annotation_type: multiselect
name: misogyny_subtypes
description: "If misogynistic, select all sub-types that apply"
labels:
- "Stereotype"
- "Shaming"
- "Objectification"
- "Violence"
tooltips:
"Stereotype": "The meme reinforces harmful gender stereotypes about women"
"Shaming": "The meme involves body-shaming, slut-shaming, or other forms of shaming women"
"Objectification": "The meme treats women as objects or reduces them to their physical appearance"
"Violence": "The meme promotes, trivializes, or jokes about violence against women"
annotation_instructions: |
You will see a meme's text overlay and a description of its image.
1. Read both the text and image description carefully.
2. Determine whether the meme is misogynistic.
3. If misogynistic, select all applicable sub-types (stereotype, shaming, objectification, violence).
Note: Consider the combination of text and image, as the misogynistic meaning may emerge from their interaction.
html_layout: |
<div style="padding: 15px; max-width: 800px; margin: auto;">
<div style="background: #fef2f2; border: 1px solid #fecaca; border-radius: 8px; padding: 16px; margin-bottom: 16px;">
<strong style="color: #991b1b;">Meme Text:</strong>
<p style="font-size: 16px; line-height: 1.7; margin: 8px 0 0 0;">{{text}}</p>
</div>
<div style="background: #fefce8; border: 1px solid #fde68a; border-radius: 8px; padding: 16px; margin-bottom: 16px;">
<strong style="color: #a16207;">Image Description:</strong>
<p style="font-size: 15px; line-height: 1.6; margin: 8px 0 0 0;">{{image_description}}</p>
</div>
</div>
allow_all_users: true
instances_per_annotator: 50
annotation_per_instance: 3
allow_skip: true
skip_reason_required: false
Sample Datasample-data.json
[
{
"id": "mami_001",
"text": "When she says she can fix things around the house",
"image_description": "Split image showing a woman holding a hammer incorrectly on the left, and a completely destroyed wall on the right with debris scattered everywhere."
},
{
"id": "mami_002",
"text": "Happy International Women's Day to all the amazing women making a difference!",
"image_description": "Collage of photos showing women in various professional roles: a scientist in a lab, a firefighter, a surgeon, and a teacher in a classroom."
}
]
// ... 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/semeval/2022/task05-multimedia-misogyny potato start config.yaml
Dataset & paper
Fersini et al., SemEval 2022
Citation (BibTeX)
@inproceedings{fersini-etal-2022-semeval,
title = "{S}em{E}val-2022 Task 5: Multimedia Automatic Misogyny Identification",
author = "Fersini, Elisabetta and Gasparini, Francesca and Rizzi, Giulia and Saibene, Aurora and Chulvi, Berta and Rosso, Paolo and Lees, Alyssa and Sorensen, Jeffrey",
booktitle = "Proceedings of the 16th International Workshop on Semantic Evaluation (SemEval-2022)",
month = jul,
year = "2022",
address = "Seattle, United States",
publisher = "Association for Computational Linguistics",
pages = "533--549",
doi = "10.18653/v1/2022.semeval-1.74",
url = "https://aclanthology.org/2022.semeval-1.74"
}Details
Annotation Types
Domain
Use Cases
Tags
Found an issue or want to improve this design?
Open an IssueRelated Designs
ADMIRE - Multimodal Idiomaticity Recognition
Multimodal idiomaticity detection task requiring annotators to identify whether expressions are used idiomatically or literally, with supporting cue analysis. Based on SemEval-2025 Task 1 (ADMIRE).
Food Hazard Detection
Food safety hazard detection task requiring annotators to identify hazards, products, and risk levels in food incident reports, and classify the type of contamination. Based on SemEval-2025 Task 9.
LLMs4Subjects - Automated Subject Tagging
Automated subject classification of academic texts, requiring annotators to assign subject categories and determine whether texts span single or multiple disciplines. Based on SemEval-2025 Task 5.