Chapter 25: NLP Projects — Spam Detection, Sentiment Analysis, Autocomplete

“Language is powerful. Teaching machines to understand it unlocks infinite possibilities.”

---

With your knowledge of tokenization, vectorization, RNNs, and Transformers, you're now ready to build end-to-end NLP applications. In this chapter, we’ll walk through three mini-projects using TensorFlow:

1. Spam Detection – Binary classification using classic vectorization
2. Sentiment Analysis – Text-to-emotion prediction using LSTM
3. Autocomplete – Next-word prediction using a Transformer

Each project includes:

• Data loading
• Text preprocessing
• Model architecture
• Training loop
• Evaluation and usage

---

1. Spam Detection with TF-IDF

Dataset: SMS Spam Collection (UCI)

import pandas as pd

df = pd.read_csv("https://raw.githubusercontent.com/justmarkham/pycon-2016-tutorial/master/data/sms.tsv", sep='\t', names=['label', 'message'])
df['label'] = df['label'].map({'ham': 0, 'spam': 1})

Preprocessing with TfidfVectorizer

from sklearn.feature_extraction.text import TfidfVectorizer
from sklearn.model_selection import train_test_split

X = TfidfVectorizer(max_features=1000).fit_transform(df['message']).toarray()
y = df['label'].values

x_train, x_test, y_train, y_test = train_test_split(X, y, test_size=0.2)

Model

from tensorflow.keras.models import Sequential
from tensorflow.keras.layers import Dense

model = Sequential([
    Dense(64, activation='relu', input_shape=(1000,)),
    Dense(1, activation='sigmoid')
])
model.compile(optimizer='adam', loss='binary_crossentropy', metrics=['accuracy'])
model.fit(x_train, y_train, epochs=5, validation_split=0.2)

---

😊 2. Sentiment Analysis with LSTM

Dataset: IMDB Reviews

(x_train, y_train), (x_test, y_test) = tf.keras.datasets.imdb.load_data(num_words=10000)
x_train = tf.keras.preprocessing.sequence.pad_sequences(x_train, maxlen=200)
x_test = tf.keras.preprocessing.sequence.pad_sequences(x_test, maxlen=200)

Model

model = tf.keras.Sequential([
    tf.keras.layers.Embedding(input_dim=10000, output_dim=64, input_length=200),
    tf.keras.layers.LSTM(64),
    tf.keras.layers.Dense(1, activation='sigmoid')
])
model.compile(optimizer='adam', loss='binary_crossentropy', metrics=['accuracy'])
model.fit(x_train, y_train, epochs=5, validation_split=0.2)

---

3. Autocomplete with Mini Transformer

This is a simplified version, using a small vocabulary.

Sample Corpus

sentences = [
    "hello how are you",
    "hello how is your day",
    "i love natural language processing",
    "tensorflow is powerful"
]

Vectorize & Build Transformer

Use TextVectorization and custom Transformer layers (as built in Chapter 24). Train it as a language model: for every input sequence, predict the next word.

---

Evaluation Tips

Project	Metric	Evaluation Example
Spam Detection	Accuracy, F1	Confusion matrix
Sentiment Analysis	Accuracy	Predict reviews manually
Autocomplete	Top-k accuracy	“hello how” → “are”, “is”, “you”

Use TensorBoard to visualize training curves, and checkpoints to save your models.

---

Deployment Ideas

Task	Deploy As
Spam Detection	FastAPI REST API
Sentiment Model	Hugging Face + Gradio
Autocomplete	TensorFlow Lite on mobile

All three can be integrated into websites, apps, or backend systems for real-time inference.

---

Summary

In this chapter, you:

• Built three full NLP projects from start to finish
• Used both traditional ML and deep learning
• Combined text preprocessing, tokenization, and modeling
• Prepared these models for deployment and real-world use

These projects are a stepping stone to larger systems: chatbots, translation, summarization, and beyond.

---