BekiChemeda/Amharic-POS-tagger-API

FastAPI backend for Amharic POS tagging with pre-trained model # 🇪🇹 Amharic POS Tagger API Hey there! Welcome to the Amharic Part-of-Speech (POS) Tagger API. This project is designed to help you analyze Amharic text by identifying the grammatical roles of each word (like Nouns, Verbs, Adjectives, etc.) using a pre-trained CRF model. I have built this with **FastAPI** to keep it snappy and modular, making it super easy for you to integrate into your own projects! --- ## 🚀 Getting Started ### 1. Set Up Your Environment First things first, make sure you have Python installed. Then, grab the dependencies: ```bash pip install -r requirements.txt ``` ### 2. Start It Up! You can start the server with a simple command: ```bash python -m app.main ``` The server will start running at `localhost`. --- ## 🛠 Features - **Blazing Fast predictions** using a tuned CRF model. - **Rate Limiting** built-in (via `slowapi`) to keep things fair and stable. - **Modular Structure** following industry best practices. - **Auto-generated Documentation**: Just head over to `/docs` once the server is running. --- ## 📖 API Documentation ### **POS Tagging Endpoint** **`POST amharic-pos-tagger-api.onre…`** Send a JSON body with the text you want to analyze: **Request:** ```json { "text": "ኢትዮጵያ ታላቅ ሀገር ናት።" } ``` **Response:** ```json { "tokens": ["ኢትዮጵያ", "ታላቅ", "ሀገር", "ናት", "።"], "pos_tags": ["N", "ADJ", "N", "AUX", "PUNC"], "tagged_sentence": "ኢትዮጵያ: N\nታላቅ: ADJ\nሀገር: N\nናት: AUX\n።: PUNC\n" } ``` --- ## 🛡 Rate Limiting To ensure everyone gets a fair share of resources, we've set a default limit of **20 requests per minute**. If you hit this limit, you'll receive a `429 Too Many Requests` response. --- ## 🤝 Contributing Feel free to open an issue or submit a pull request if you have ideas to make this even better! Made with ❤️ for the Ethiopian NLP community.