Aspect-Based Sentiment Analysis for Turkish

Capstone project developing a multi-label sentiment analysis system for Turkish language using BERT

📋 Project Information

Institution: T.C. Maltepe University
Faculty: Engineering and Natural Sciences
Department: Software Engineering
Type: Capstone Project

Team Members:

• Mustafa Tayyip BAYRAM - LinkedIn
• Furkan ÖCALAN - LinkedIn

Advisor: Assist. Prof. Dr. Volkan TUNALI

📂 Structure

• Analyze/ - Core analysis engine and model integration
• Models/ - Classification models
• Documents/ - Project reports and documentation
• Semeval-Analysis/ - Dataset analysis scripts
• apps/ - Flask application modules
• migrations/ - Database migrations
• uploads/ - User file uploads directory

🎯 Project Overview

A web-based application that performs aspect-based sentiment analysis on Turkish text using a fine-tuned BERT model. The system analyzes text to identify specific aspects and their associated sentiments.

🔬 Technical Implementation

Dataset

• Source: SemEval 2016 Turkish dataset
• Task: Multi-label aspect-based sentiment classification

Preprocessing Pipeline

• Zemberek: Turkish language processing
• NLTK: Natural language toolkit for text processing

Model Architecture

• Base Model: Pre-trained BERT
• Task: Multi-label classification
• Performance:
  • Accuracy: 0.69
  • Recall: 0.65

Web Application Stack

Application Modules

• Authentication: User registration and login
• File: Upload and process text files
• Home: Main analysis interface
• Templates: Rendering components
• Config: Database configuration

🚀 Installation & Setup

1. Install Dependencies

pip install -r requirements.txt

2. Model Setup

Download the trained BERT model and place it in Analyze/Models/ directory.
Note: Model size restrictions prevent direct upload. Contact team for access.

3. Database Configuration

Update database credentials in apps/config.py (line 13)

4. Run Application

# Development mode
$env:FLASK_ENV="development"
$env:FLASK_DEBUG=1
$env:FLASK_APP=".\run.py"
flask run

📊 Application Screenshots


🛠️ Development Tools

📚 Features

• Turkish language sentiment analysis
• Aspect-based opinion mining
• Multi-label classification
• Interactive data visualization
• File-based batch processing
• User authentication system
• Results export functionality

📖 Documentation

Complete project reports and documentation are available in the Documents folder.

🔗 Related Resources

Similar app: Credit Card Fraud Detection Repository

---

Natural language processing application for Turkish sentiment analysis