temsa/OpenMed-PPSN-v5_1

OpenMed PPSN v5.1

Token classification model derived from OpenMed/OpenMed-PII-SuperClinical-Large-434M-v1 with B-PPSN / I-PPSN support for Irish PPSN detection.

Why v5.1

This iteration hardens PPSN behavior against known false positives on number-like strings (for example phone numbers and malformed ID-like tokens) while preserving non-PPSN behavior.

What this release contains

• —Full model weights (model.safetensors) with original OpenMed labels + PPSN labels.
• —label_meta.json with label mapping and provenance.
• —Eval artifacts:
• —eval_manual_irish_v5_1_large_v2_raw.json
• —eval_hybrid_v5_1_large_v2_strict.json
• —ab_non_ppsn_v5_1.json
• —qa_ppsn_regression_v6_validated.jsonl
• —eval_manual_qa_regression_v6_validated_v5_1_raw.json
• —eval_hybrid_qa_regression_v6_validated_v5_1_strict.json

Key results

On irish_ppsn_eval_large_v2:

• —Raw model-only PPSN performance:
• —P=0.8699 R=0.9956 F1=0.9285
• —Recommended strict hybrid mode (--no-plausible-ppsn, --ppsn-min-score 0.6):
• —P=1.0000 R=0.9997 F1=0.9999

Non-PPSN retention vs base OpenMed (ab_non_ppsn_v5_1.json):

• —Synthetic non-PPSN F1 delta vs base: +0.00049
• —Real-set agreement F1 (candidate vs base): 1.0000
• —Real entities per 1k chars delta vs base: 0.0000

Recommended production usage

Use strict hybrid PPSN post-processing (checksum-backed) for production masking. Raw model-only PPSN spans are less reliable on malformed numeric strings.

For a quick local smoke test of the packaged checkpoint, use the bundled word_aligned helper:

python3 inference_word_aligned.py \
  --ppsn-min-score 0.4 \
  --text "My PPSN is 1234567TW and I need help with my housing grant." \
  --json

Install dependencies from pyproject.toml first if you are not already in an environment with transformers, torch, and regex.

License and attribution

• —This derivative release is distributed under Apache-2.0, consistent with the base model license tag.
• —Base model: OpenMed/OpenMed-PII-SuperClinical-Large-434M-v1.
• —See NOTICE for attribution details.

<!-- portfolio-comparison:start -->

Portfolio Comparison

Updated: 2026-03-16.

Use this section for the fastest public comparison across the temsa PII masking portfolio.

• —The first core table only includes public checkpoints that ship both comparable q8 accuracy and q8 CPU throughput.
• —The first PPSN table only includes public artifacts that ship comparable PPSN accuracy and CPU throughput.
• —Missing cells in the archive tables mean the older release did not ship that metric in its public bundle.
• —DiffMask rows use the reconciled clean_single_pass harness that matches the deployed runtime.
• —GlobalPointer rows use the public raw-only span-matrix release bundle and its packaged q8 ONNX artifact.
• —The same content is shipped as PORTFOLIO_COMPARISON.md inside each public model repo.

Irish Core PII: Comparable Public Checkpoints

Irish Core PII: Other Public Checkpoints

Finance-boundary q8 F1 is 1.0000 for OpenMed-mLiteClinical-IrishCorePII-135M-v2-rc6, OpenMed-mLiteClinical-IrishCorePII-135M-v2-rc7, OpenMed-mLiteClinical-IrishCorePII-135M-v2-rc8, and all public IrishCore-DiffMask releases from rc1 to rc6. OpenMed-mLiteClinical-IrishCorePII-135M-v2-rc5 ships 0.8750 on that public q8 suite.

PPSN-Only: Comparable Public Artifacts

PPSN-Only: Historical Public Checkpoints

If you need the strongest current raw-only Irish core model, start with IrishCore-GlobalPointer-135M-v1-rc4. If you need the fastest CPU-first raw-only line, compare it against IrishCore-DiffMask-135M-v1-rc6. If you need a PPSN-only artifact, compare the canonical fp32, fp16, and q8 variants of OpenMed-mLiteClinical-IrishPPSN-135M-v1 directly in the table above. <!-- portfolio-comparison:end -->