ner-agent

A simple, language-agnostic Named Entity Recognition (NER) agent powered by LLMs.

Features

• Multilingual: Supports English, Chinese, Japanese, Korean, French, German, Russian, Spanish, and more.
• Entity Types: Recognizes PERSON, NORP (nationalities, religions, political groups, languages), LOCATION, DATETIME, NUMERIC, and PROPER_NOUN (events, works, organizations, products, etc.).
• Easy Integration: Use as a Python library with a simple async API.
• Tested: Includes comprehensive test cases for multiple languages and entity types.

Installation

pip install ner-agent

Or clone and install locally:

git clone https://github.com/allen2c/ner-agent.git
cd ner-agent
pip install .

Python 3.11+ required.

Usage

import asyncio
from ner_agent import NerAgent

async def main():
    agent = NerAgent()
    text = "Elon Musk visited Tesla's Gigafactory in Austin on March 15, 2024, announcing a 20% increase."
    result = await agent.run(text)
    for entity in result.entities:
        print(entity)

asyncio.run(main())

Output:

name='PERSON' value='Elon Musk' start=0 end=9
name='PROPER_NOUN' value='Tesla' start=18 end=23
name='LOCATION' value='Gigafactory' start=26 end=37
...

Entity Types

• PERSON: People, including fictional characters.
• NORP: Nationalities, religious groups, political groups, languages.
• LOCATION: Geopolitical entities, facilities, places.
• DATETIME: Dates, times, periods, ages.
• NUMERIC: Numbers, money, quantities, percentages, ordinals/cardinals.
• PROPER_NOUN: Named events, works, laws, products, organizations, companies, etc.

Testing

To run the tests:

pytest

Configuration

• By default, uses OpenAI-compatible LLMs via openai-agents.
• You can configure the model and OpenAI client (see tests/conftest.py for examples).

License

MIT License

For more details, see ner_agent/init.py and the tests.

If you need more advanced usage or want to contribute, please check the GitHub repository.