Spam detection pipeline in NLP - Model Pipeline Trace

Model Pipeline - Spam detection pipeline

This spam detection pipeline takes email text and decides if it is spam or not. It cleans the text, turns words into numbers, trains a model to learn patterns, and then predicts new emails as spam or not.

Data Flow - 7 Stages

1Raw Email Text

1000 emails x 1 column (text)→Collect raw email messages as text→1000 emails x 1 column (text)

"Win a free phone now!"

↓

2Text Cleaning

1000 emails x 1 column (text)→Lowercase, remove punctuation and stopwords→1000 emails x 1 column (cleaned text)

"win free phone"

↓

3Feature Extraction

1000 emails x 1 column (cleaned text)→Convert words to numbers using TF-IDF vectorizer→1000 emails x 5000 columns (features)

[0, 0, 1.2, 0, ..., 0.5]

↓

4Train/Test Split

1000 emails x 5000 features→Split data into 800 training and 200 testing emails→Train: 800 x 5000, Test: 200 x 5000

Train features shape: 800 x 5000

↓

5Model Training

Train: 800 x 5000 features→Train logistic regression model on training data→Trained model

Model learns weights for features

↓

6Model Evaluation

Test: 200 x 5000 features→Predict on test data and calculate accuracy→Accuracy score (e.g., 0.92)

Model predicts 184 correct out of 200

↓

7Prediction

New email text→Clean text, extract features, predict spam or not→Spam label (0 = not spam, 1 = spam)

"Congratulations, you won!" -> 1

Training Trace - Epoch by Epoch

Loss
0.5 |****
0.4 |*** 
0.3 |**  
0.2 |*   
0.1 |    
     1 2 3 4 5 Epochs

Epoch	Loss ↓	Accuracy ↑	Observation
1	0.45	0.78	Model starts learning basic spam patterns
2	0.32	0.85	Loss decreases, accuracy improves
3	0.25	0.89	Model captures more features
4	0.20	0.91	Training stabilizes with good accuracy
5	0.18	0.92	Final epoch with best performance

Prediction Trace - 4 Layers

Layer 1: Input Email Text

Layer 2: TF-IDF Vectorizer

Layer 3: Logistic Regression Model

Layer 4: Thresholding

Model Quiz - 3 Questions

What happens to the email text during the 'Text Cleaning' stage?

A It is converted into a probability score

B It is lowercased and punctuation is removed

C It is split into training and testing sets

D It is labeled as spam or not spam

Key Insight

This visualization shows how text data is transformed step-by-step into numbers that a model can understand. The model learns to spot spam by reducing errors over time, improving accuracy. Finally, it predicts new emails as spam or not based on learned patterns.