Clickbait Title Classification
A textual dataset composed of article titles annotated as “clickbait” or not, designed to train automatic detection models.
Description
The dataset Clickbait Title Classification contains 32,000 article titles taken from various sites, some of which are known for publishing “catchy” content. Each title is annotated as clickbait (1) or non-clickbait (0). It is derived from a research project on the automatic detection of clickbait in online media.
What is this dataset for?
- Train a classification model for short texts (binary)
- Analyzing the linguistic strategies of clickbait in the online press
- Evaluate or refine automatic detection models (e.g. for moderation, SEO, etc.)
Can it be enriched or improved?
Yes. It is possible to cross it with contextual data (source, date, click rate) or to refine the annotations (degree of clickbait, types of formulations, emotion aroused). Multilingual variants can also be added for international projects.
🔎 In summary
🧠 Recommended for
- NLP students
- AI engineers working on moderation or editorial content
- Researchers in digital communication
🔧 Compatible tools
- Scikit-learn
- Pandas
- TensorFlow
- Hugging Face Transformers
💡 Tip
To improve the robustness of the model, cross this dataset with real-time data (RSS, social networks) and enrich with contextual embeddings such as BERT.
Frequently Asked Questions
Does this dataset contain the sources of the titles analyzed?
No, only sentences are provided with their annotation, but some typical sources (BuzzFeed, etc.) are mentioned in the associated study.
Can it be used to detect misleading titles in French?
Not directly, but the structure can be used as a basis for creating an equivalent multilingual dataset or translating titles for fine-tuning.
Is this dataset suitable for a classification project with few resources?
Yes, it is an excellent starting point for lightweight projects with classical or compressed models (e.g. logistic regression, Distilbert).
