feat(tasks 2 & 3): Add deep learning and NLP scripts to src/

Includes the TensorFlow CNN script for MNIST and the spaCy script for NER and Sentiment Analysis.

Author: augusto-mate

src/task2_deep_learning_cnn.py

@@ -0,0 +1,119 @@
+# -*- coding: utf-8 -*-
+#
+# Task 2: Deep Learning with TensorFlow/Keras
+# Objective: Build and train a CNN to classify handwritten digits (MNIST)
+# Framework: TensorFlow / Keras (Deep Learning)
+
+import tensorflow as tf
+from tensorflow.keras.datasets import mnist
+from tensorflow.keras.models import Sequential
+from tensorflow.keras.layers import Conv2D, MaxPooling2D, Flatten, Dense, Dropout
+from tensorflow.keras.utils import to_categorical
+import numpy as np
+import matplotlib.pyplot as plt
+
+# Ensure TensorFlow runs only once for logging 
+print("--- Task 2: CNN Image Classification (MNIST Dataset) ---")
+
+# 1. Load and Preprocess Data 
+try:
+    (x_train, y_train), (x_test, y_test) = mnist.load_data()
+    
+    # Reshape: Add channel dimension (28x28 -> 28x28x1)
+    x_train = x_train.reshape(x_train.shape[0], 28, 28, 1)
+    x_test = x_test.reshape(x_test.shape[0], 28, 28, 1)
+
+    # Normalize: Scale pixel values from 0-255 to 0-1 
+    x_train = x_train.astype('float32') / 255
+    x_test = x_test.astype('float32') / 255
+    
+    # One-Hot Encoding: Convert labels to binary vectors 
+    num_classes = 10
+    y_train_encoded = to_categorical(y_train, num_classes)
+    y_test_encoded = to_categorical(y_test, num_classes)
+
+    print(f"\nTraining data shape: {x_train.shape}")
+
+except Exception as e:
+    print(f"Error loading or preprocessing MNIST data: {e}")
+    exit()
+
+
+# 2. Define the CNN Model Architecture 
+model = Sequential([
+    # Input layer and first Convolution 
+    Conv2D(32, kernel_size=(3, 3), activation='relu', input_shape=(28, 28, 1)),
+    MaxPooling2D(pool_size=(2, 2)),
+    
+    # Second Convolution
+    Conv2D(64, (3, 3), activation='relu'),
+    MaxPooling2D(pool_size=(2, 2)),
+    
+    Dropout(0.25),
+    
+    # Fully Connected Layers 
+    Flatten(),
+    Dense(128, activation='relu'),
+    Dropout(0.5),
+    
+    # Output Layer 
+    Dense(num_classes, activation='softmax')
+])
+
+# 3. Compile the Model 
+print("\n3. Compiling Model...")
+model.compile(
+    optimizer='adam',
+    loss='categorical_crossentropy', # Loss function for multi-class classification 
+    metrics=['accuracy']
+)
+
+model.summary()
+
+
+# 4. Train the Model 
+print("\n4. Training the CNN Model...")
+history = model.fit(
+    x_train, y_train_encoded,
+    batch_size=128,
+    epochs=10, # 10 epochs is sufficient for high accuracy 
+    verbose=1,
+    validation_data=(x_test, y_test_encoded)
+)
+print("   Training finished.")
+
+
+# 5. Evaluate Performance 
+print("\n5. Evaluating Model Performance on Test Set...")
+score = model.evaluate(x_test, y_test_encoded, verbose=0)
+print(f"Test Loss: {score[0]:.4f}")
+print(f"Test Accuracy: {score[1]:.4f}")
+
+# Check if the target accuracy is met 
+if score[1] > 0.95:
+    print("   Success! Test Accuracy target (>95%) achieved.")
+else:
+    print("   Note: Test Accuracy target not reached. Review model parameters.")
+
+
+# 6. Visualize Predictions 
+test_samples = x_test[:5]
+test_labels = y_test[:5]
+predictions = model.predict(test_samples)
+predicted_classes = np.argmax(predictions, axis=1)
+
+print("\n6. Visualizing 5 Sample Predictions (will require manual display of the plot).")
+
+# Plotting the results 
+plt.figure(figsize=(12, 4))
+for i in range(5):
+    plt.subplot(1, 5, i + 1)
+    plt.imshow(test_samples[i].reshape(28, 28), cmap='gray')
+    plt.title(f"True: {test_labels[i]}\nPred: {predicted_classes[i]}", fontsize=10)
+    plt.axis('off')
+plt.suptitle("CNN Predictions on Sample MNIST Images")
+# In a real environment, you'd save this image for the report 
+# plt.savefig('assets/cnn_predictions_sample.png')
+# plt.show() # Uncomment to display plot
+
+print("\nTask 2 completed. CNN trained and evaluated.")

src/task3_nlp_spacy.py

@@ -0,0 +1,88 @@
+# -*- coding: utf-8 -*-
+#
+# Task 3: NLP with spaCy
+# Objective: Perform Named Entity Recognition (NER) and rule-based Sentiment Analysis on user reviews.
+# Framework: spaCy (Natural Language Processing)
+
+import spacy
+from spacy.lang.en.stop_words import STOP_WORDS
+
+# Note: The model 'en_core_web_sm' must be downloaded separately 
+# (Ex: python -m spacy download en_core_web_sm)
+
+print("--- Task 3: NLP - NER and Rule-based Sentiment ---")
+
+# 1. Sample Text Data (Reviews)
+
+amazon_reviews = [
+    "The new 'ChronoWatch X1' is absolutely fantastic! Battery life is great, and the 'Ogetec' brand delivered quickly.",
+    "I was disappointed with the 'ZenBook 14'. It overheats constantly. I regret buying this laptop.",
+    "This book, 'The AI Engineer', is highly informative and a must-read for any developer. Five stars!",
+    "The customer service from 'EcoGoods' was terrible. My package arrived late and damaged."
+]
+
+# 2. Load spaCy Model 
+try:
+    # Load the English model 
+    nlp = spacy.load("en_core_web_sm")
+    print("\nspaCy model 'en_core_web_sm' loaded successfully.")
+except OSError:
+    print("\n[ERROR] spaCy model 'en_core_web_sm' not found.")
+    print("Please run: python -m spacy download en_core_web_sm")
+    exit()
+
+# 3. Named Entity Recognition (NER) 
+print("\n3. Named Entity Recognition (NER) Results:")
+
+def perform_ner(text):
+    """
+    Applies the spaCy model to extract named entities.
+    """
+    doc = nlp(text)
+    entities = [(ent.text, ent.label_) for ent in doc.ents]
+    return entities
+
+for i, review in enumerate(amazon_reviews):
+    entities = perform_ner(review)
+    
+    # Filter for entities relevant to products/brands 
+    product_brands = [(text, label) for text, label in entities if label in ('ORG', 'PRODUCT', 'WORK_OF_ART')]
+    
+    print(f"\nReview {i+1}: '{review[:50]}...'")
+    if product_brands:
+        print(f"  Extracted Entities (Product/Brand): {product_brands}")
+    else:
+        print("  No relevant entities found.")
+
+
+# 4. Rule-Based Sentiment Analysis 
+# This is a simple, rule-based approach using keywords, as required by the task.
+
+positive_words = {"fantastic", "great", "highly informative", "must-read", "five stars", "excellent", "love"}
+negative_words = {"disappointed", "overheats", "regret", "terrible", "late", "damaged", "awful"}
+
+def analyze_sentiment(text):
+    """
+    Performs basic sentiment analysis by counting positive and negative keywords.
+    """
+    doc = nlp(text.lower())
+    
+    # Tokenization and removing stop words 
+    tokens = [token.text for token in doc if token.text not in STOP_WORDS and token.is_alpha]
+    
+    pos_count = sum(1 for token in tokens if token in positive_words)
+    neg_count = sum(1 for token in tokens if token in negative_words)
+    
+    if pos_count > neg_count:
+        return "Positive"
+    elif neg_count > pos_count:
+        return "Negative"
+    else:
+        return "Neutral/Mixed"
+
+print("\n4. Rule-Based Sentiment Analysis Results:")
+for i, review in enumerate(amazon_reviews):
+    sentiment = analyze_sentiment(review)
+    print(f"Review {i+1}: Sentiment: {sentiment}")
+
+print("\nTask 3 completed successfully. NER and Sentiment analysis performed.")