Review Analyzer

A comprehensive tool for analyzing game reviews using AI-powered sentiment and aspect analysis. The application processes raw review data, extracts sentiment and aspects from individual sentences, labels aspects, and generates detailed analysis reports with visualizations.

Overview

The Review Analyzer is designed to process large volumes of game reviews to extract meaningful insights. It performs three main operations:

1. Sentence Processing: Extracts sentiment and aspects from individual review sentences
2. Aspect Labeling: Labels and categorizes different aspects of the game (gameplay, graphics, story, etc.)
3. Analysis & Visualization: Generates comprehensive analysis reports with charts and statistics

Features

• Multi-threaded processing for efficient handling of large datasets
• AI-powered analysis using Mistral language model
• Comprehensive logging with both console and file output
• Automatic chart generation for visual analysis
• Flexible configuration with command-line arguments
• Structured output in multiple formats (JSON, CSV, charts)

Installation

Prerequisites

• Python 3.13
• Ollama installed and running locally

Setup

1. Clone the repository

   git clone https://github.com/DFGANDP/ReviewAnalyzer.git
   cd SteamReviewAnalyzer

2. Install dependencies

   pip install -r requirements.txt

3. Install and start Ollama

   # Follow instructions at https://ollama.ai/
   ollama pull MHKetbi/Mistral-Small3.1-24B-Instruct-2503:q5_K_L

4. Prepare input data

   • Place your review JSON file in the review_analyzer/input/ directory
   • The default expected file is 105600_20250209173825.json

Usage

Basic Usage

Run the analysis from the project root directory:

python -m review_analyzer.presentation.main

Command Line Arguments

Argument	Type	Default	Description
--workers	int	6	Number of worker threads for parallel processing
--language	str	"english"	Language for review processing
--limit	int	None	Limit the number of reviews to process (useful for testing)

Examples

Process with custom worker count:

python -m review_analyzer.presentation.main --workers 8

Process reviews in a different language:

python -m review_analyzer.presentation.main --language polish

Process only first 100 reviews (for testing):

python -m review_analyzer.presentation.main --limit 100

Combine multiple arguments:

python -m review_analyzer.presentation.main --workers 4 --language english --limit 50

Project Structure

SteamReviewAnalyzer/
├── review_analyzer/
│   ├── config.py                 # Configuration and paths
│   ├── domain/                   # Domain models and interfaces
│   │   ├── aspect_labeler.py
│   │   ├── interfaces.py
│   │   └── models.py
│   ├── infrastructure/           # Data access and external services
│   │   ├── aspect_labeler.py    # AI aspect labeling
│   │   ├── dataframe_loader.py  # CSV data loading
│   │   ├── dataframe_saver.py   # CSV data saving
│   │   ├── global_analyzer.py   # Analysis engine
│   │   ├── json_loader.py       # JSON data loading
│   │   ├── json_saver.py        # JSON data saving
│   │   ├── log_handlers/        # Logging configuration
│   │   ├── mistral_extractor.py # AI sentiment extraction
│   │   ├── sentence_analyzer.py # Sentence processing
│   │   ├── sentence_loader.py   # Sentence data loading
│   │   └── utils.py             # Utility functions
│   ├── input/                   # Input data directory
│   ├── output/                  # Generated output (timestamped)
│   ├── presentation/            # Application entry points
│   │   ├── main.py             # Main CLI interface
│   │   ├── runner.py           # Core execution logic
│   │   └── [other runners]     # Specialized runners
│   ├── prompts/                # AI prompt templates
│   │   ├── prompt_extract.txt  # Sentence extraction prompts
│   │   └── prompt_label.txt    # Aspect labeling prompts
│   └── service/                # Business logic services
│       ├── aspect_labeling_service.py
│       └── review_sentence_processing_service.py
├── tests/                      # Test suite
├── sandbox/                    # Development and testing scripts
├── requirements.txt            # Python dependencies
└── README.md                  # This file

Output Structure

The application generates timestamped output directories containing:

• Analysis reports (analysis_liked.json, analysis_disliked.json)
• Data files (CSV files with labeled aspects)
• Charts (PNG visualizations for each aspect)
• Logs (Detailed processing logs)

Dependencies

• numpy: Numerical computing
• pandas: Data manipulation and analysis
• matplotlib: Chart generation
• seaborn: Statistical data visualization
• ollama: Local AI model interface
• tqdm: Progress bars
• pytest: Testing framework

Development

Running Tests

pytest tests/

Adding New Features

1. Follow the existing architecture patterns
2. Add tests for new functionality
3. Update documentation as needed

Troubleshooting

Common Issues:

1. Ollama not running: Ensure Ollama is installed and the model is pulled
2. Memory issues: Reduce --workers parameter for large datasets
3. Input file not found: Verify the JSON file exists in the input directory

Example of charts