Universal Ner Ita

🚀 Universal NER for Italian (Zero-Shot)

This is a universal model for Italian Named Entity Recognition (NER) using zero-shot learning, applicable across all domains.

🚀 Quick Start

It's crucial to understand that this model is universal and works across all domains. If you're aiming for an F1 score close to 90/99% for a specific domain, feel free to contact Michele Montebovi via email at montebovi.michele@gmail.com. This way, the model can be customized to better meet your Italian entity recognition needs.

You can try the model here: https://huggingface.co/spaces/DeepMount00/universal_ner_ita

✨ Features

• Universal Applicability: This model is designed to operate across all domains, providing flexibility for various entity recognition tasks in Italian.
• Zero-Shot Learning: It uses a zero-shot learning approach, which allows it to identify a wide range of entities without specific training on those entities.

📦 Installation

To use this model, you need to download the GLiNER project:

!pip install gliner

💻 Usage Examples

Basic Usage

from gliner import GLiNER

model = GLiNER.from_pretrained("DeepMount00/universal_ner_ita")

text = """
Il comune di Castelrosso, con codice fiscale 80012345678, ha approvato il finanziamento di 15.000€ destinati alla ristrutturazione del parco giochi cittadino, affidando l'incarico alla società 'Verde Vivo Società Cooperativa', con sede legale in Corso della Libertà 45, Verona, da completarsi entro il 30/09/2024.
"""

labels = ["comune", "codice fiscale", "importo", "società", "indirizzo", "data di completamento"]

entities = model.predict_entities(text, labels)

max_length = max(len(entity["text"]) for entity in entities)

for entity in entities:
    padded_text = entity["text"].ljust(max_length)
    print(f"{padded_text} => {entity['label']}")

📚 Documentation

Model Description

This model is intended for Named Entity Recognition (NER) tasks, specifically for the Italian language. It adopts a zero-shot learning approach, enabling it to recognize a wide variety of entities without the need for specific training on those entities. This makes it highly adaptable for various applications that require entity extraction from Italian text.

Model Performance

• Inference Time: The model runs on CPUs, and it has an inference time of 0.01 seconds on a GPU. The performance on a CPU will vary depending on the specific hardware configuration.

Try It Out

You can test the model directly in your browser through the following Hugging Face Spaces link: https://huggingface.co/spaces/DeepMount00/universal_ner_ita.

📄 License

This model is licensed under the Apache-2.0 license.