Fine-grained sentiment (5-class) in NLP - Model Pipeline Trace

Model Pipeline - Fine-grained sentiment (5-class)

This pipeline reads text reviews and predicts one of five sentiment classes: very negative, negative, neutral, positive, or very positive. It cleans the text, converts words into numbers, trains a model to learn patterns, and then predicts sentiment labels.

Data Flow - 5 Stages

1Raw Text Input

1000 rows x 1 column→Load 1000 text reviews→1000 rows x 1 column

"The movie was fantastic and thrilling!"

↓

2Text Cleaning

1000 rows x 1 column→Lowercase, remove punctuation, and strip extra spaces→1000 rows x 1 column

"the movie was fantastic and thrilling"

↓

3Tokenization & Padding

1000 rows x 1 column→Convert words to sequences of integers and pad to length 20→1000 rows x 20 columns

[12, 45, 78, 9, 34, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0]

↓

4Train/Test Split

1000 rows x 20 columns→Split data into 800 training and 200 testing samples→Train: 800 rows x 20 columns, Test: 200 rows x 20 columns

Train sample: [12, 45, 78, ...], Test sample: [7, 23, 56, ...]

↓

5Model Training

Train: 800 rows x 20 columns→Train embedding + LSTM + dense layers to classify into 5 sentiment classes→Model with 5-class output layer

Model learns to map sequences to sentiment labels 0-4

Training Trace - Epoch by Epoch

Loss: 1.45 |****
       1.20 |****
       1.05 |***
       0.95 |**
       0.88 |*
Epochs -> 1    2    3    4    5

Epoch	Loss ↓	Accuracy ↑	Observation
1	1.45	0.35	Model starts learning with low accuracy and high loss
2	1.20	0.48	Loss decreases and accuracy improves as model learns
3	1.05	0.56	Model continues to improve steadily
4	0.95	0.62	Training converges with better predictions
5	0.88	0.67	Final epoch shows good accuracy and low loss

Prediction Trace - 5 Layers

Layer 1: Input Text

Layer 2: Tokenization & Padding

Layer 3: Embedding Layer

Layer 4: LSTM Layer

Layer 5: Dense + Softmax Output

Model Quiz - 3 Questions

What does the tokenization step do in this pipeline?

A Splits data into training and testing sets

B Removes stop words from the text

C Converts words into numbers for the model

D Calculates accuracy during training

Key Insight

This visualization shows how text data is transformed step-by-step into numbers that a model can understand. The model learns to predict one of five sentiment classes by improving accuracy and reducing loss over training. Softmax outputs give clear probabilities for each sentiment, helping us understand the model's confidence.