pdfviz private-information detector source record
=================================================

Recorded: 2026-09-02

Release package
---------------

Public path: /models/pii-93777d70/
Package format: ort-gzip-parts-v1
Manifest SHA-256: 93777d70b1108bc74d5077db454da87af8ce774f352072d9a004665e99f2e886
ORT model SHA-256: f8336da49694fa6a30136b089b0663e8b808ed11b2b3114bc7d134a390d58e74
ONNX model SHA-256: 7bcd6c3049d737885cd7e3c0eee3ffafd358c67f2a186c7d1f4c863648dd0778
FP32 fine-tuned weights SHA-256: 74b5982cba603156956c76f625b8f328f82ff45e17dd9f6a507c1993a25afbb1

The detector is a pdfviz fine-tune and quantization of Google BERT Mini. It
runs with ONNX Runtime Web in a browser worker. It is not a hosted inference
service.

Base model
----------

Repository: https://huggingface.co/google/bert_uncased_L-4_H-256_A-4
Upstream source: https://github.com/google-research/bert
Licence: Apache-2.0
Licence text: /licenses/pdfjs-dist.txt

Runtime
-------

Project: ONNX Runtime Web 1.29.0
Source: https://github.com/microsoft/onnxruntime/tree/v1.29.0/js/web
Licence: MIT
Licence text: /licenses/onnxruntime.txt

Inherited v8 training sources
-----------------------------

Dataiku Kiji PII training data
Repository: https://huggingface.co/datasets/DataikuNLP/kiji-pii-training-data
Revision: 0275550f0b1f1b8f2dc9356fd31ac1c788b8228b
Licence: Apache-2.0

E3-JSI Synthetic Multi PII NER v1
Repository: https://huggingface.co/datasets/E3-JSI/synthetic-multi-pii-ner-v1
Revision: 2fe1b17ea2d82d80b2570e5c48c416a3b317828f
Licence: MIT

V9 training and evaluation sources
----------------------------------

Gretel PII Masking English v1
Repository: https://huggingface.co/datasets/gretelai/gretel-pii-masking-en-v1
Revision: e06eb1499ca8d54470f085021cd8e54f9efac7fd
Licence: Apache-2.0

Text Anonymization Benchmark
Repository: https://github.com/NorskRegnesentral/text-anonymization-benchmark
Dataset revision: 558e09e26d6b36f5f78440074e6a233946d98bd9
Licence: MIT

RedactionBench was used only after model and threshold freeze as test evidence.
No RedactionBench row was used for training or tuning.
Repository: https://huggingface.co/datasets/RedactionBench/RedactionBench
Revision: d45e9cec89bc49c69355e252fec29cc0229982f6
Test artifact SHA-256: 17ea0b577344917ce6e265667dd833cbf18e4f2cc07aa230d55f1e151219f5f0
Licence: CC-BY-4.0

Text Anonymization Benchmark MIT notice
---------------------------------------

The MIT License (MIT)

Copyright (C) 2021-2026 Norsk Regnesentral

Permission is hereby granted, free of charge, to any person obtaining a copy of
this software and associated documentation files (the "Software"), to deal in
the Software without restriction, including without limitation the rights to
use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies
of the Software, and to permit persons to whom the Software is furnished to do
so, subject to the following conditions:

The above copyright notice and this permission notice shall be included in all
copies or substantial portions of the Software.

THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
SOFTWARE.