One-hot encoding for text in NLP - ML Experiment: Train & Evaluate

Experiment - One-hot encoding for text

Problem:You want to convert a list of text sentences into one-hot encoded vectors to prepare data for a machine learning model.

Current Metrics:N/A - currently, the text is raw and not encoded, so the model cannot train.

Issue:The text data is not in a numeric format that machine learning models can understand. Without encoding, the model cannot learn from the text.

Your Task

Convert the given list of sentences into one-hot encoded vectors using a vocabulary built from the text. Verify the encoding by printing the one-hot vectors.

Use only Python standard libraries and scikit-learn.

Do not use embedding layers or other complex encodings.

Keep the vocabulary size manageable by using the unique words from the input sentences.

Hint 1

Hint 2

Hint 3

Solution

NLP

from sklearn.feature_extraction.text import CountVectorizer

# Sample text data
sentences = [
    "I love machine learning",
    "Machine learning is fun",
    "I love coding",
    "Coding is fun"
]

# Create CountVectorizer with binary=True for one-hot encoding
vectorizer = CountVectorizer(binary=True)

# Fit on sentences to build vocabulary
vectorizer.fit(sentences)

# Transform sentences to one-hot encoded vectors
one_hot_vectors = vectorizer.transform(sentences)

# Convert sparse matrix to array for display
one_hot_array = one_hot_vectors.toarray()

# Print vocabulary and one-hot vectors
print("Vocabulary:", vectorizer.get_feature_names_out())
print("One-hot encoded vectors:")
for sentence, vector in zip(sentences, one_hot_array):
    print(f"{sentence}: {vector}")

Used CountVectorizer with binary=True to create one-hot encoding.

Built vocabulary from the input sentences.

Transformed sentences into one-hot encoded vectors.

Printed vocabulary and vectors to verify encoding.

Results Interpretation

Before encoding, the text was raw and unusable by ML models.

After encoding, each sentence is represented as a vector of 0s and 1s indicating presence of words.

This numeric format can now be fed into ML models.

One-hot encoding converts text into a simple numeric format by marking which words appear in each sentence. This is a basic but important step to prepare text data for machine learning.

Bonus Experiment

Try using TF-IDF encoding instead of one-hot encoding to see how word importance affects the vectors.

💡 Hint

Use sklearn's TfidfVectorizer and compare the output vectors to one-hot vectors.