🚀 Clickbait Headline Detection System
An end-to-end NLP classification model built using DistilBERT to accurately identify whether a given headline or title is Clickbait or Not Clickbait.
👤 Author & Student Metadata
- Developer / Author: RAJA SAAD ALI
- Roll Number: SU72-BSCSM-F25-011
- Program / Degree: BS Computer Science (BSCS)
- Project Title: Clickbait Headline Detection System
🛠️ Model Architecture & Training Setup
- Base Architecture:
distilbert-base-uncased - Task Type: Binary Sequence Classification
- Frameworks: PyTorch, Hugging Face
transformers,datasets - Max Sequence Length: 128 tokens
- Optimizer: AdamW
- Output Classes:
LABEL_0: NOT CLICKBAIT (Genuine / Standard News)LABEL_1: CLICKBAIT (Sensationalized / Engagement Bait)
📊 Dataset & Evaluation Metrics
- Dataset: Clickbait Detection Dataset (Balanced News Titles and Social Media Headlines)
- Evaluation Focus: High Accuracy, Precision, and Recall across both real and clickbait samples.
💻 How to Use in Python
Option 1: Quick Pipeline Inference
from transformers import pipeline
# Load classifier
classifier = pipeline("text-classification", model="Saman11233/clickbait-distilbert-detector")
# Test sample
sample_headline = "10 Secrets Doctors Don't Want You to Know!"
result = classifier(sample_headline)
print(result)
Option 2: Interactive Testing Loop (Google Colab)
from transformers import pipeline
classifier = pipeline("text-classification", model="Saman11233/clickbait-distilbert-detector")
user_input = input("Enter headline: ")
if user_input.strip():
res = classifier(user_input)[0]
label = "CLICKBAIT" if res["label"] == "LABEL_1" else "NOT CLICKBAIT"
print(f"Prediction: {label} ({res['score']*100:.2f}% confidence)")
📌 Intended Use & Limitations
- Intended Use: Automating content moderation, news feed curation, and assisting users in filtering clickbait content.
- Limitations: Performs best on English news titles and short social media posts.