Finguys/acta-anonymizer-financial

Acta Anonymizer Financial Adapter

This model is a fine-tuned adapter for Romanian financial text anonymization. It's based on XLM-RoBERTa and trained specifically for detecting and anonymizing PII in Romanian financial documents from Moldova.

Key features:

• Romanian language support
• Financial domain specialization
• GDPR compliance focused
• High accuracy PII detection

Use cases:

• Banking document anonymization
• Financial report processing
• Compliance data handling

Current Version: 20250914_103007

Key Features

• Romanian language support
• GDPR compliance focused
• High accuracy PII detection
• Domain-specific fine-tuning

Use Cases

• Banking document anonymization
• Financial report processing
• Compliance data handling

Training Data

This model was trained on synthetic Moldovan PII data for financial domain anonymization.

Usage

from peft import PeftModel
from transformers import AutoModelForTokenClassification, AutoTokenizer, pipeline

# Load base model
model = AutoModelForTokenClassification.from_pretrained("EvanD/xlm-roberta-base-romanian-ner-ronec")
tokenizer = AutoTokenizer.from_pretrained("EvanD/xlm-roberta-base-romanian-ner-ronec")

# Load adapter
model = PeftModel.from_pretrained(model, "Finguys/acta-anonymizer-financial")

# Create pipeline
ner_pipeline = pipeline(
    "token-classification",
    model=model,
    tokenizer=tokenizer,
    aggregation_strategy="simple"
)

# Example usage
text = "Ion Popescu are un cont la Banca Transilvania cu IBAN RO49AAAA1B310075938400000."
entities = ner_pipeline(text)
print(entities)

Training

This model was trained using LoRA (Low-Rank Adaptation) on synthetic Moldovan PII data.

Versions

• Latest: Root level contains the most recent version
• Archived: Previous versions are stored in versions/ folder
• Version Index: See version_history.yaml for complete version history