Skip to content
Showcase/Hate Speech Detection
intermediatetext

Hate Speech Detection

Identify and categorize hate speech, offensive language, and toxic content in text.

Select all that apply:

Configuration Fileconfig.yaml

yaml
# Hate Speech Detection Configuration
# Identify and categorize toxic content

task_dir: "."
annotation_task_name: "Hate Speech Detection"

data_files:
  - "data/posts.json"

item_properties:
  id_key: "id"
  text_key: "text"
  text_display_key: "text"

user_config:
  allow_all_users: true

annotation_schemes:
  - annotation_type: "radio"
    name: "primary_label"
    description: "How would you classify this content?"
    labels:
      - name: "Hate Speech"
        tooltip: "Attacks a group based on protected characteristics"
        key_value: "h"
        color: "#dc2626"
      - name: "Offensive Language"
        tooltip: "Vulgar or inappropriate but not targeting a group"
        key_value: "o"
        color: "#f97316"
      - name: "Neither"
        tooltip: "No hate speech or offensive language"
        key_value: "n"
        color: "#22c55e"

  - annotation_type: "multiselect"
    name: "target_groups"
    description: "If hate speech, which groups are targeted?"
    labels:
      - name: "Race/Ethnicity"
      - name: "Religion"
      - name: "Gender"
      - name: "Sexual Orientation"
      - name: "Disability"
      - name: "National Origin"
      - name: "Age"
      - name: "Other"
    show_if:
      field: "primary_label"
      value: "Hate Speech"

  - annotation_type: "multiselect"
    name: "hate_type"
    description: "What type of hate speech?"
    labels:
      - name: "Slurs/Epithets"
      - name: "Dehumanization"
      - name: "Stereotyping"
      - name: "Threats/Incitement"
      - name: "Exclusion/Marginalization"
    show_if:
      field: "primary_label"
      value: "Hate Speech"

  - annotation_type: "likert"
    name: "severity"
    description: "How severe is this content?"
    size: 5
    min_label: "Mild"
    max_label: "Severe"

  - annotation_type: "radio"
    name: "action_recommendation"
    description: "Recommended action"
    labels:
      - name: "No action"
      - name: "Warning label"
      - name: "Hide behind warning"
      - name: "Remove content"
      - name: "Remove + ban user"

output: "annotation_output/"

output_annotation_dir: "annotation_output/"
output_annotation_format: "json"

Sample Datasample-data.json

json
[
  {
    "id": "hs_001",
    "text": "I can't believe how long the line was at the DMV today. Absolutely ridiculous!",
    "source": "social_media",
    "timestamp": "2024-01-15T10:30:00Z"
  },
  {
    "id": "hs_002",
    "text": "That movie was so stupid, I want my two hours back.",
    "source": "social_media",
    "timestamp": "2024-01-15T11:45:00Z"
  }
]

// ... and 1 more items

Try it live — no install

Boot the real Potato server in your browser (WebAssembly) and annotate with this exact config. Nothing leaves your machine.

▶ Run live in your browser

Get This Design

View on GitHub

Clone or download from the repository

Quick start:

git clone https://github.com/davidjurgens/potato-showcase.git
cd potato-showcase/templates/text/hate-speech-detection
potato start config.yaml

Details

Annotation Types

radiomultiselect

Domain

nlpcontent-moderation

Use Cases

content-moderationtoxicity-detection

Tags

hate-speechtoxicitymoderationoffensiveclassification

Found an issue or want to improve this design?

Open an Issue

Related Designs

Intent Classification

Classify user utterances into intents for chatbot and virtual assistant training.

radiomultiselect

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).

radiomultiselect

AfriHate - Hate and Abusive Language for African Languages

Multilingual content-moderation annotation following the AfriHate scheme (Muhammad et al., NAACL 2025), a collection of hate speech and abusive language datasets for 15 African languages including Amharic, Hausa, Igbo, Yoruba, Swahili, Somali and isiZulu. Native speakers familiar with the regional culture assign each tweet a three-way label - hate, abusive, or neutral - and, when the tweet is hateful, mark the targeted attribute (e.g. ethnicity, religion, politics, gender). Sample items are mild, constructed illustrations that avoid real slurs or targeting of real groups; an English gloss is provided for reference only.

radiomultiselect