🤖 Aspect-Based Sentiment Analysis (ABSA) API

A high-performance REST API for Aspect-Based Sentiment Analysis that processes both individual texts and large-scale CSV datasets using fine-tuned BERT models.

Features • Quick Start • Model Setup • API Docs

---

🚀 Features

• 🧠 Fine-Tuned BERT trained on MAMS dataset + SpaCy for aspect extraction
• ⚡ Async Processing with Celery workers for bulk CSV analysis
• 🔐 JWT Authentication for secure API access
• 📊 Unified Session System handles both sync text and async file uploads
• 📑 Swagger UI for interactive API testing

---

🏗️ Architecture

graph LR
    A[Client] -->|Text/CSV| B[Django API]
    B -->|Sync| C[BERT Model]
    B -->|Async| D[Redis Queue]
    D --> E[Celery Worker]
    E --> C
    C -->|Results| F[(Database)]
    F --> A

Loading

Workflow: Text requests → instant analysis | CSV uploads → background processing with progress tracking

---

🛠️ Tech Stack

Layer	Technology
API	Django REST Framework
Queue	Celery + Redis
ML	PyTorch + Transformers + SpaCy
Auth	JWT (simplejwt)
DB	SQLite (dev) / PostgreSQL (prod)

---

⚡ Quick Start

Prerequisites

• Python 3.11+
• Redis 7.x+
• Git

Installation

# 1. Clone and navigate
git clone https://github.com/yourusername/absa-api.git
cd absa-api

# 2. Setup virtual environment
python -m venv venv
source venv/bin/activate  # On Windows: venv\Scripts\activate

# 3. Install dependencies
pip install -r requirements.txt
python -m spacy download en_core_web_sm

# 4. Setup database
python manage.py migrate
python manage.py createsuperuser

# 5. Start Redis (in separate terminal)
redis-server

Run the Application

Open 3 terminals:

# Terminal 1 - API Server
python manage.py runserver

# Terminal 2 - Celery Worker
## Windows
celery -A ABSAapi worker --pool=solo -l info
## Mac/Linux
celery -A ABSAapi worker -l info

# Terminal 3 - Celery Beat (optional)
celery -A ABSAapi beat -l info

✅ API Ready: http://127.0.0.1:8000/api
📖 Docs: http://127.0.0.1:8000/api/docs/

---

🧠 Model Setup

The fine-tuned BERT model is required for sentiment analysis. Choose one option:

Option 1: Download From Kaggle (Easiest way)

Download Model

Option 2: Train Locally (Recommended for customization)

• Download the dataset

• Run train Notebook

    jupyter notebook Notebooks/train.ipynb

• Fine-tune BERT for aspect-based sentiment analysis

• Save the model to ./models/bert_absa/

Option 3: Train on Kaggle (For more details and GPU access)

1. Open the Kaggle notebook ABSA BERT Training and run it
2. Enable GPU accelerator in notebook settings
3. Run all cells to train the model
4. Download the trained model from Kaggle output
5. Place in ./models/bert_absa/ directory

Model Directory Structure:

models/
└── bert_absa/
    ├── config.json
    ├── pytorch_model.bin
    └── vocab.txt

---

📖 API Documentation

Interactive Docs: http://127.0.0.1:8000/api/docs/

Quick Reference

Endpoint	Method	Description	Auth
/api/register	POST	Register a new user	❌
/api/login/	POST	Get JWT Bearer Token (for code/apps)	❌
/api-auth/login/	GET	Browser Login (Standard HTML Page)	❌
/api/sessions/	POST	Analyze text/CSV	✅
/api/sessions/	GET	List sessions	✅
/api/sessions/{id}/	GET	Get results	✅
/api/sessions/{id}/	GET	Get results	✅

---

⭐ Star this repo if you find it useful!