dheerubhadoria/English-Toxic-Language-Dataset-for-NLP

English Toxic Language Dataset for NLP

The English Toxic Language Dataset for NLP is a synthetic dataset containing 10,000 labeled English comments designed for toxic language detection, content moderation, and natural language processing (NLP) research. It is suitable for developers, data scientists, and researchers building AI systems that can identify and classify toxic, abusive, or harmful text.

The dataset includes binary labels (Clean and Toxic), toxicity categories, severity scores, and language information, making it easy to train and evaluate text classification models using TensorFlow, PyTorch, Scikit-learn, Hugging Face Transformers, and TensorFlow Lite.

Dataset Features

• 10,000 English comments

• Binary classification (Clean / Toxic)

• Toxicity categories

  • Abuse
  • Insult
  • Harassment
  • Hate
  • Threat
  • Profanity
  • Spam

• Severity score for each comment

• English language

• CSV format

• Ready for machine learning and deep learning workflows

Dataset Schema

Column	Description
id	Unique sample identifier
text	English comment
label	0 = Clean, 1 = Toxic
category	Toxicity category
severity	Toxicity severity score
language	Language code (en)

Use Cases

• Toxic comment detection
• AI content moderation
• Chatbot safety
• Social media moderation
• Offensive language detection
• NLP research
• Text classification
• TensorFlow Lite mobile applications
• AI safety experiments
• Educational machine learning projects

Supported Frameworks

• TensorFlow
• TensorFlow Lite
• PyTorch
• Hugging Face Transformers
• Scikit-learn
• Keras

Keywords

Toxic Language Detection, Toxic Comment Classification, NLP Dataset, Content Moderation, Machine Learning, Deep Learning, AI Safety, TensorFlow Dataset, PyTorch Dataset, English Text Dataset, Chat Moderation, Offensive Language Detection, Binary Text Classification.

Important Note

This dataset is synthetically generated for research, experimentation, educational purposes, and prototype development. It is intended to support the development and evaluation of machine learning models for toxic language detection and should not be considered a replacement for large-scale, human-annotated moderation datasets.