NER-based PII Recognition using ONNX Runtime

This crate provides Named Entity Recognition (NER) capabilities for PII detection using quantized ONNX models for efficient inference.

Features

• ONNX Runtime integration for model inference
• Support for quantized int8 models
• Token-based NER with entity span detection
• Compatible with various NER model architectures (BERT, RoBERTa, etc.)

Example

use redact_ner::NerRecognizer;
use redact_core::recognizers::Recognizer;

// Load model
let recognizer = NerRecognizer::from_file("model.onnx").unwrap();

// Analyze text
let text = "John Doe works at Acme Corp in New York";
let results = recognizer.analyze(text, "en").unwrap();

for result in results {
    println!("{:?}: {}", result.entity_type, result.text.unwrap());
}