Sentiment Analysis with Naive Bayes in C

Overview

This project implements a Naive Bayes classifier in C to perform sentiment analysis on Amazon reviews. It processes textual data, constructs a vocabulary using a hashmap, and applies probabilistic calculations to classify reviews as positive or negative.

Features

• Efficient text preprocessing with tokenization and stop-word filtering.
• Hashmap-based vocabulary storage for fast lookups.
• Probabilistic sentiment classification using Naive Bayes.
• Logging support for debugging and analyzing results.
• 67.19% accuracy on the test dataset.

Requirements

• C Compiler (e.g., GCC)
• Libraries:
  • cJSON for JSON parsing
  • Standard C libraries for string manipulation and file handling

How It Works

1. Data Preprocessing: Reviews are tokenized, cleaned, and stored in a hashmap.
2. Model Training: The vocabulary is populated with word frequencies for both positive and negative classes.
3. Classification: The model calculates probabilities using Naive Bayes and logs classification results.
4. Performance Metrics: Reports accuracy, failures, and zero-error cases for evaluation.

Dataset Information

The dataset is not included in the repository. Instead, you can access the dataset here.

@article{hou2024bridging,
  title={Bridging Language and Items for Retrieval and Recommendation},
  author={Hou, Yupeng and Li, Jiacheng and He, Zhankui and Yan, An and Chen, Xiusi and McAuley, Julian},
  journal={arXiv preprint arXiv:2403.03952},
  year={2024}
}