nel-models

Named Entity Linking Models

Trained spaCy models for named entity linking (NEL) tasks, focusing on company disambiguation and entity resolution.

📦 Available Models

Company Named Entity Linking

Disambiguate company mentions in text by linking them to a knowledge base of known entities.

Model	Size	Description	Download
en_company_nel_lg	~500MB	Large model with better accuracy	Latest Release
en_company_nel_md	~100MB	Medium model, faster inference	Latest Release

🚀 Installation

From GitHub Releases (Recommended)

# Install the large model
pip install https://github.com/sondalex/nel-models/releases/download/en_company_nel_lg-0.1.0/en_company_nel_lg-0.1.0-py3-none-any.whl

# Install the medium model
pip install https://github.com/sondalex/nel-models/releases/download/en_company_nel_md-0.1.0/en_company_nel_md-0.1.0-py3-none-any.whl

💻 Usage

import spacy

# Load the model
nlp = spacy.load("en_company_nel_lg")

# Process text
text = "Apple announced record profits.  Microsoft also reported strong earnings."
doc = nlp(text)

# Access linked entities
for ent in doc. ents:
    print(f"{ent.text} -> {ent.kb_id_} ({ent.label_})")

Training Data

Models are trained on a combination of:

1. FNSPID Dataset: Financial news

   • License: CC BY 4.0 equivalent
   • Authors: Zihan Dong, Xinyu Fan, Zhiyuan Peng

2. Company Entities Dataset: Curated knowledge base of companies and aliases

   • License: CC BY-SA 4.0 & CC0 1.0
   • Author: Alexandre Sonderegger

📝 License

See individual model metadata files for detailed license information.

Attribution

If you use the company entity linking models in your research or projects, please cite:

@software{sonderegger2025companynel,
  author = {Sonderegger, Alexandre},
  title = {Company Named Entity Linking Models for spaCy},
  year = {2025},
  url = {https://github.com/sondalex/nel-models},
  version = {0.1.0}
}