https://github.com/mahdizynali/deep-analysis-persian-news
Deep context explicit and text analysis of Persian news with deep NLP and using Google Trax and jax and (seprated pytorch branch)
https://github.com/mahdizynali/deep-analysis-persian-news
deep-mind machine-learning natural-language-processing nlp persian-nlp pytorch tensorflow text-analysis trax
Last synced: 4 months ago
JSON representation
Deep context explicit and text analysis of Persian news with deep NLP and using Google Trax and jax and (seprated pytorch branch)
- Host: GitHub
- URL: https://github.com/mahdizynali/deep-analysis-persian-news
- Owner: mahdizynali
- Created: 2025-06-12T06:56:47.000Z (5 months ago)
- Default Branch: main
- Last Pushed: 2025-07-13T20:09:11.000Z (4 months ago)
- Last Synced: 2025-07-13T21:26:04.034Z (4 months ago)
- Topics: deep-mind, machine-learning, natural-language-processing, nlp, persian-nlp, pytorch, tensorflow, text-analysis, trax
- Language: Python
- Homepage:
- Size: 37.1 KB
- Stars: 1
- Watchers: 0
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
README
# Deep Analysis of Persian News 📰🧠🇮🇷
This project performs deep natural language analysis on Persian news articles using modern neural network architectures. It leverages an LSTM-based model built with [Trax](https://github.com/google/trax) and [JAX](https://github.com/google/jax) to understand and classify the explicit content of Persian news text. Also i have provided seprated Pytorch branch for those whom prefers traditional way. Final model has been traiend on over 6 Millions parameters.
## 🔍 Project Goals
- Perform **context-aware analysis** on Persian news headlines or full articles
- Leverage **LSTM-based deep learning models** for semantic understanding
- Achieve strong performance benchmarks on large-scale Persian datasets
## 🚀 Key Features
- 🧠 **Deep Neural Network**: LSTM-based model implemented using Trax and JAX
- 📰 **Persian Language Support**: Tailored preprocessing and custom tokenization for Persian Lang
- 📊 **Large Dataset**: Trained on over **200,000 news articles**
- ✅ **High Accuracy**: Achieves **80% classification accuracy** on validation set
## Benchmark
On first try with a simple Trax network.
| Dataset | Loss | Accuracy | Parameters |
|------------|------|----------|------------|
| Train | 0.60223989 | 0.84375000 | 6 Millions |
| Validation | 0.65386836 | 0.80412000 | 6 Millions |
---
## 📦 Try It
1. Clone the repository:
```bash
git clone https://github.com/mahdizynali/Deep-Analysis-Persian-News.git
cd Deep-Analysis-Persian-News
```
```
pip3 install -r requirements.txt
```
For train a new model :
```
python3 main.py train
```
Try prediction:
```
python3 main.py predict
```