Tamil Lemmatizer

Tamil Lemmatizer is a character-level lemmatization library for Tamil text.
It normalizes inflected Tamil word forms and maps them to their base lemma using a deep learning model (PyTorch).

✨ Features

• ✅ Lemmatizes Tamil words to their base form
• ✅ Handles unseen words using a character-level sequence model
• ✅ Simple Python API
• ✅ Supports batch inference
• ✅ Open-source and extensible

📦 Installation

pip install tamil-lemmatizer

🚀 Quick Start

from tamil_lemmatizer import TamilLemmatizer

lemmatizer = TamilLemmatizer()

word = "சென்றார்கள்"
lemma = lemmatizer.lemmatize(word)

print(lemma)   # Output: செல்

Batch input

words = ["பாடுகிறது", "வந்தார்கள்", "சென்றேன்"]
print(lemmatizer.lemmatize_batch(words))

📚 Description

Tamil is morphologically rich. A single lemma can have hundreds of inflected variations. This library uses:

• A character-level encoder-decoder architecture
• Trained using PyTorch on a curated Tamil lemma dataset
• Supports lemmatization for verbs and nouns

🛠️ Model Architecture

• Encoder: BiLSTM or Transformer (depending on version)
• Decoder: Attention-based sequence generator
• Loss: Cross entropy over Tamil character vocabulary

🔧 CLI Usage

tamil-lemmatizer "வந்தார்கள்"

📄 License

This project is released under the MIT License.

🤝 Contributing

Pull requests are welcome. If contributing major changes, open an issue first to discuss what you want to change.

✉️ Contact

Maintainer: Hemanth Kumar GitHub: Hemanth Thunder