Clasificación de Sentimientos en Reseñas de Películas con Naive Bayes en Python

2026-02-20 · Source: Naturallanguageprocessing on Medium · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Data Science & Analytics · Depth: Intermediate, short

Summary

A sentiment classification system was developed using the NLTK `movie_reviews` dataset and the Multinomial Naive Bayes algorithm. The system classifies 2000 movie reviews (1000 positive, 1000 negative) as either positive or negative. Text preprocessing involved lowercasing, number and punctuation removal, tokenization, and stopword removal, including domain-specific terms like "film" and "movie." Text was numerically represented using `CountVectorizer` (Bag-of-Words). The model achieved an accuracy of 0.8175 and an F1-score of 0.81 on a 80/20 train-test split. Interpretability analysis identified discriminative words, such as "excellent" for positive sentiment and "terrible" for negative, by examining `log P(word | class)` differences.

Key takeaway

For Data Scientists building text classification systems, understanding the impact of preprocessing and model interpretability is crucial. Your choice of a simple model like Naive Bayes can yield solid results (e.g., 82% F1-score) if the data is balanced and features are well-engineered. Focus on identifying truly discriminative words rather than just frequent ones to explain model decisions and guide feature refinement.

Key insights

Naive Bayes, with proper preprocessing, offers robust and interpretable sentiment classification.

Principles

• Balanced datasets prevent classification bias.
• Preprocessing enhances text classification quality.
• Discriminative words reveal model decision-making.

Method

Develop a binary sentiment classifier by preprocessing text (lowercasing, tokenization, stopword removal), vectorizing with `CountVectorizer`, and training a Multinomial Naive Bayes model on a balanced dataset.

In practice

• Extend stopword lists with domain-specific terms.
• Use `feature_log_prob_` for Naive Bayes interpretability.
• Consider TF-IDF or n-grams for future improvements.

Topics

• Sentiment Analysis
• Naive Bayes
• Natural Language Processing
• Text Preprocessing
• Movie Review Classification

Best for: Machine Learning Engineer, Data Scientist, AI Student

Related on AIssential

• Sentiment Analysis with Naive Bayes: A Step-by-Step NLP Walkthrough — A sentiment analysis pipeline uses Naive Bayes, demonstrating text preprocessing and Bag of Words feature extraction.
• Natural Language Processing: A Beginner’s Guide from Someone Who’s Learning It Too — NLP transforms unstructured human language into computer-understandable data, powering diverse AI applications.
• From Logistic Regression to GPT-2: Building a Complete Spam Detection & Sentiment Analysis Pipeline — For imbalanced text classification, prioritize F1-score and Precision-Recall AUC over accuracy and ROC-AUC.
• On GATE, Text and Social Media Analysis, and Detecting Misinformation Online — A multi-layer analytical pipeline effectively extracts and categorizes digital identity aspects of agrifood products from online discourse.
• From Text to Numbers: Demystifying Text Preprocessing in NLP — NLP text preprocessing, using methods like Bag of Words and TF-IDF, translates human language into numerical data for machine learning models.
• Exploration of symbolic methods for emotion detection for Portuguese — Symbolic emotion detection in Portuguese shows stable relative performance despite domain-specific absolute variations.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Naturallanguageprocessing on Medium.