Aspect-Based Sentiment Analysis (ABSA) for Restaurant Reviews

Project Overview

This repository contains an implementation of Aspect-Based Sentiment Analysis (ABSA) for restaurant reviews. The project aims to analyze restaurant reviews at a fine-grained level, extracting sentiments about specific aspects of the restaurant experience (food quality, service, ambiance, etc.) rather than just the overall sentiment of the review.

Features

• Multiple Model Implementations:
  • Transformer-based ABSA models using DistilBERT
  • Baseline sentiment models (Naive Bayes, SVM)
  • Performance comparison between approaches
• Comprehensive Dataset:
  • Restaurant reviews with aspect-level annotations
  • Train/Dev/Test splits for robust evaluation
  • Cleaned and preprocessed versions of datasets
• Evaluation Framework:
  • Accuracy, F1, and other relevant metrics
  • Performance visualization tools
  • Detailed analysis of model results

Repository Structure

dsa-absa/
├── Dataset/               # Restaurant reviews dataset with ABSA annotations
├── dataset_stats/         # Visualizations of dataset characteristics
├── evaluation/            # Evaluation scripts and metrics
├── logs/                  # Training logs
├── models/                # Model implementations
│   ├── absa/              # Transformer-based ABSA models
│   ├── baseline/          # Naive Bayes baseline models
│   └── svm/               # SVM baseline models
├── results/               # Model checkpoints and evaluation results
├── saved_model/           # Best performing models
├── scripts/               # Utility scripts for visualization and analysis
├── training_plots/        # Training performance visualizations
└── utils/                 # Utility functions for data processing

Dataset

The project uses a restaurant review dataset with aspect-based annotations in the format:

• tokens: Tokenized review text
• absa1, absa2, absa3: Aspect annotations including:
  • Position indices of aspect terms
  • Aspect category (e.g., FOOD#QUALITY, SERVICE#GENERAL)
  • Sentiment polarity (0=negative, 1=neutral, 2=positive)

Model Architecture

The primary ABSA model uses DistilBERT, a lightweight transformer model, fine-tuned for aspect-based sentiment classification. The architecture includes:

• DistilBERT encoder for text representation
• Classification head for sentiment prediction
• Custom data preprocessing for aspect extraction

Setup and Installation

# Clone the repository
git clone https://github.com/yourusername/dsa-absa.git
cd dsa-absa

# Install dependencies
pip install -r requirements.txt

# Download required NLTK data and spaCy model
python -m nltk.downloader punkt
python -m spacy download en_core_web_sm

Usage

Training the ABSA Model

cd models/absa
python absa-1.py

Training Baseline Models

cd models/baseline
python train_baseline.py

# OR for SVM baseline
cd models/svm
python train_svm_baseline.py

Evaluation

cd evaluation
python evaluate_metrics.py

Visualizing Results

cd scripts
python plot_training_stats.py

Training all models and Creating Images/Chart

python evaluation/compare_models.py

Note: This command must be executed from the project root to ensure proper access to datasets and internal modules.

Results

The transformer-based ABSA model achieves superior performance compared to baseline models:

Model	Accuracy	F1 Score
DistilBERT ABSA	~0.85	~0.84
Naive Bayes	~0.70	~0.68
SVM	~0.75	~0.74

License

This project is licensed under the terms of the LICENSE file included in the repository.

Acknowledgments

• This project was developed as a final project for the Machine Learning course.
• The dataset is based on the SemEval-2016 restaurant reviews dataset.