🌍 Language Detection App using SimpleRNN

📌 Project Overview

This project is an end-to-end Language Detection application built using SimpleRNN and deployed with Streamlit. It predicts the language of a given text among 17 languages such as English, Hindi, French, Malayalam, Spanish, Tamil, Arabic, etc.

Project Highlights:

• Multi-class text classification
• Handling class imbalance
• Complete NLP pipeline (training → evaluation → deployment)
• Simple and interpretable RNN-based model

---

🧠 Model Architecture

The model is kept simple to focus on core concepts.

Architecture:

• Embedding Layer
• SimpleRNN Layer
• Dropout Layer
• Dense Softmax Output Layer

Flow: Embedding → SimpleRNN → Dropout → Dense (Softmax)

Why SimpleRNN?

• Easy to explain (interview-friendly)
• Captures sequential text patterns
• Lightweight and fast for deployment

---

📊 Dataset Information

• Source: Kaggle – Language Detection Dataset

• Total Languages: 17

• Total Samples: 10,267

• Columns:

  • Text (input sentence)
  • Language (target label)

⚠️ Dataset is imbalanced, so class weights are used.

---

🚀 Model Performance

Metric	Value
Test Accuracy	95.8%
Test Loss	0.16

Improved performance due to:

• Class weighting
• Early stopping
• Proper preprocessing

---

🛠️ Tech Stack

• Python
• TensorFlow / Keras
• Scikit-learn
• Streamlit
• NumPy, Pandas, Matplotlib, Seaborn

---

📁 Project Structure

Language-Detection-RNN/
├── saved_model/
│   ├── simple_rnn_model.h5
│   └── tokenizer.pkl
├── eda.ipynb
├── prediction.ipynb
├── app.py
├── requirements.txt
└── readme.md

---

⚙️ Installation Steps

1️⃣ Clone the Repository

git clone https://github.com/Adi3042/Language-Detection-using-RNN.git
cd Language-Detection-using-RNN

2️⃣ (Optional) Create Virtual Environment

python -m venv venv
source venv/bin/activate        # Linux/Mac
venv\Scripts\activate           # Windows

3️⃣ Install Dependencies

pip install -r requirements.txt

---

▶️ Run the Project

streamlit run app.py

---

🧪 Example

Input:

यह एक अच्छा दिन है

Output:

Predicted Language: Hindi
Confidence: 0.98

---