Fake News or Real News?

Enter any news headline and content below to get an instant verdict.

REAL NEWS
🧠 Deep Learning Model Architecture

The binary text classifier follows a 6-stage deep learning pipeline using Long Short-Term Memory (LSTM) per the project specification:

1
Input News Article
Combines Headline + Article Body Text into unified content stream
String Text
â–¼
2
NLP Cleaning & Tokenization
Lowercasing, punctuation removal, word ID mapping & sequence padding
maxlen=300
â–¼
3
Embedding Layer
Converts word integer IDs into dense 128-dimensional semantic vectors
dim=128
â–¼
4
LSTM Layer (Long Short-Term Memory)
Learns sequential word order dependencies and linguistic context
units=64
â–¼
5
Dropout Regularization
Randomly zeroes 30% of activations to prevent neural overfitting
rate=0.3
â–¼
6
Dense Layer (Sigmoid Activation)
Computes final binary probability: < 0.5 = Fake News, ≥ 0.5 = Real News
Sigmoid (0 or 1)
Layer Type Output Shape / Parameters Core Function
1 Embedding (vocab: 20000, dim: 128, mask_zero=True) Converts word IDs into dense semantic vectors
2 LSTM (units: 64, return_sequences=False) Learns sequential context & dependencies
3 Dropout (rate: 0.3) Regularization layer reducing overfitting
4 Dense (units: 1, activation='sigmoid') Outputs binary classification score (0 or 1)