teimurjan/tanaos-text-anonymizer-onnx

Text Anonymizer NER (ONNX)

Named Entity Recognition model for PII detection. Quantized int8 ONNX format optimized for browser inference.

Model Details

• Architecture: RoBERTa-based token classification
• Format: ONNX (int8 quantized, ~119MB)
• Original Model: tanaos/tanaos-text-anonymizer-v1

Detected Entities

Entity	Description
PERSON	Names of people
LOCATION	Geographic locations
ADDRESS	Street addresses
PHONE_NUMBER	Phone numbers
DATE	Dates and time expressions

Usage with transformers.js

import { pipeline } from '@huggingface/transformers';

const ner = await pipeline(
  'token-classification',
  'teimurjan/tanaos-text-anonymizer-onnx',
  { device: 'wasm' }
);

const result = await ner(
  'John Doe lives at 123 Main St, New York. His phone is (555) 123-4567.',
  { aggregation_strategy: 'first' }
);
console.log(result);
// [
//   { entity_group: 'PERSON', word: 'John Doe', score: 0.99 },
//   { entity_group: 'ADDRESS', word: '123 Main St', score: 0.98 },
//   { entity_group: 'LOCATION', word: 'New York', score: 0.97 },
//   { entity_group: 'PHONE_NUMBER', word: '(555) 123-4567', score: 0.99 }
// ]

License

Apache 2.0 (following the original model license)