GolemPII v2: Hebrew PII detection

Demo: try it Blog: read the write-up GitHub: code and examples Parameters: 45M REDACT F1: 0.884 Language: Hebrew Licence: CC BY 4.0 Cite: BibTeX Author: liranbaba.dev

GolemPII v2 tiny

Version 2.0, 2026-11-13.

The smallest variant of GolemPII-v2-base: the same 15 labels and the same training data on dicta-il/dictabert-tiny (45M parameters). It is about five times faster than the main model and less accurate on enterprise text. Use it when speed or size matters more than accuracy; otherwise use the main model.

The labels, the training data, the way the scores are measured and the limits are described in the main model's card. The numbers below are from the REDACT final half, which was scored once after training.

Which model should I use

Model Parameters Use it for
GolemPII-v2-base 277M Hebrew mixed with English in enterprise text (emails, tickets, logs, JSON), and the best recall of the three on identity, card and bank numbers
GolemPII-v2-small 125M Hebrew prose: the best of the three at telling person names from organisations, at less than half the size of base and the same speed
GolemPII-v2-tiny (this model) 45M When speed or size matters most (about five times faster), with lower accuracy on enterprise text

Usage

from transformers import pipeline

detector = pipeline(
    "token-classification",
    model="CordwainerSmith/GolemPII-v2-tiny",
    aggregation_strategy="simple",
    stride=128,  # lets the pipeline handle texts longer than 512 tokens
)

TEXT = """נושא: RE: ticket #48213 - Χ’Χ“Χ›Χ•ΧŸ Χ€Χ¨Χ˜Χ™ ΧœΧ§Χ•Χ—

Χ”Χ™Χ™ Χ¨Χ•Χ Χ™Χͺ,
Χ”ΧœΧ§Χ•Χ— יואב Χ‘ΧŸ Χ“Χ•Χ“ (Χͺ.Χ–. 034567891) Χ”ΧͺΧ§Χ©Χ¨ Χ”Χ‘Χ•Χ§Χ¨ Χ•Χ‘Χ™Χ§Χ© ΧœΧ’Χ“Χ›ΧŸ אΧͺ Χ”Χ›ΧͺΧ•Χ‘Χͺ ΧœΧžΧ©ΧœΧ•Χ—: Χ¨Χ—Χ•Χ‘ Χ”Χ¨Χ¦Χœ 42, רמΧͺ Χ’ΧŸ, ΧžΧ™Χ§Χ•Χ“ 5244302.
א׀שר ΧœΧ”Χ©Χ™Χ’ אוΧͺΧ• Χ‘-054-7731920 או Χ‘ΧžΧ™Χ™Χœ yoav.bendavid@example.com.
Χ”Χ—Χ™Χ•Χ‘ Χ”Χ—Χ•Χ“Χ©Χ™ Χ™Χ¨Χ“ ΧžΧ›Χ¨Χ˜Χ™Χ‘ Χ•Χ™Χ–Χ” 4580 1234 5678 9015, Χ•Χ”Χ–Χ™Χ›Χ•Χ™ Χ™Χ•Χ’Χ‘Χ¨ ΧœΧ—Χ©Χ‘Χ•ΧŸ IL620108000000099999999.
order_id=ORD-2024-88127, בכום: 349.90 Χ©"Χ—, ΧͺΧΧ¨Χ™Χš ΧœΧ™Χ“Χ” ΧœΧΧ™ΧžΧ•Χͺ: 14/03/1987.

ΧͺΧ•Χ“Χ”,
Dana Levi | Customer Success"""

EDGE_PUNCTUATION = " .,;:!?()[]{}\"'-"


def clean_spans(text, entities):
    """Trim punctuation from span edges and join spans of one label that touch or overlap."""
    spans = []
    for entity in entities:
        start, end, label = entity["start"], entity["end"], entity["entity_group"]
        while start < end and text[start] in EDGE_PUNCTUATION:
            start += 1
        while end > start and text[end - 1] in EDGE_PUNCTUATION:
            end -= 1
        if start == end:
            continue
        if spans and spans[-1][2] == label and start <= spans[-1][1]:
            spans[-1] = (spans[-1][0], max(end, spans[-1][1]), label)
        else:
            spans.append((start, end, label))
    return spans


def redact(text, spans):
    pieces, position = [], 0
    for start, end, label in spans:
        pieces += [text[position:start], f"[{label}]"]
        position = end
    return "".join(pieces) + text[position:]


spans = clean_spans(TEXT, detector(TEXT))
for start, end, label in spans:
    print(f"{label:<12} {TEXT[start:end]}")
print(redact(TEXT, spans))

The detected spans:

FIRST_NAME   Χ¨Χ•Χ Χ™Χͺ
FIRST_NAME   יואב
LAST_NAME    Χ‘ΧŸ Χ“Χ•Χ“
ID_NUM       034567891
STREET       Χ¨Χ—Χ•Χ‘ Χ”Χ¨Χ¦Χœ 42
CITY         רמΧͺ Χ’ΧŸ
POSTAL_CODE  5244302
PHONE_NUM    054-7731920
EMAIL        yoav.bendavid@example.com
CC_PROVIDER  Χ•Χ™Χ–Χ”
CC_NUM       4580 1234 5678 9015
IBAN         IL620108000000099999999
DATE         14/03/1987
FIRST_NAME   Dana
LAST_NAME    Levi

The redacted text. The ticket number, the order id and the amount are left alone:

נושא: RE: ticket #48213 - Χ’Χ“Χ›Χ•ΧŸ Χ€Χ¨Χ˜Χ™ ΧœΧ§Χ•Χ—

Χ”Χ™Χ™ [FIRST_NAME],
Χ”ΧœΧ§Χ•Χ— [FIRST_NAME] [LAST_NAME] (Χͺ.Χ–. [ID_NUM]) Χ”ΧͺΧ§Χ©Χ¨ Χ”Χ‘Χ•Χ§Χ¨ Χ•Χ‘Χ™Χ§Χ© ΧœΧ’Χ“Χ›ΧŸ אΧͺ Χ”Χ›ΧͺΧ•Χ‘Χͺ ΧœΧžΧ©ΧœΧ•Χ—: [STREET], [CITY], ΧžΧ™Χ§Χ•Χ“ [POSTAL_CODE].
א׀שר ΧœΧ”Χ©Χ™Χ’ אוΧͺΧ• Χ‘-[PHONE_NUM] או Χ‘ΧžΧ™Χ™Χœ [EMAIL].
Χ”Χ—Χ™Χ•Χ‘ Χ”Χ—Χ•Χ“Χ©Χ™ Χ™Χ¨Χ“ ΧžΧ›Χ¨Χ˜Χ™Χ‘ [CC_PROVIDER] [CC_NUM], Χ•Χ”Χ–Χ™Χ›Χ•Χ™ Χ™Χ•Χ’Χ‘Χ¨ ΧœΧ—Χ©Χ‘Χ•ΧŸ [IBAN].
order_id=ORD-2024-88127, בכום: 349.90 Χ©"Χ—, ΧͺΧΧ¨Χ™Χš ΧœΧ™Χ“Χ” ΧœΧΧ™ΧžΧ•Χͺ: [DATE].

ΧͺΧ•Χ“Χ”,
[FIRST_NAME] [LAST_NAME] | Customer Success

clean_spans is optional. The pipeline groups sub-word tokens, so a raw span can carry a bracket, a hyphen or a full stop at its edge, or arrive in two pieces; the function trims and joins them. The scores on this page were measured on the raw pipeline output, without it.

The encoder is uncased.

With Microsoft Presidio

A ready recognizer for Presidio is in the GitHub repo. It registers this model for language he, maps the labels to Presidio entity names, and works with the standard analyze and anonymize calls.

Results

The evaluation code published with GolemPII-v2-base produced every number below, and scores any token-classification model on the same sets.

REDACT final, five categories

Model Parameters Character P / R / F1 Entity P / R / F1
GolemPII-v1 277M 0.592 / 0.884 / 0.709 0.601 / 0.913 / 0.725
GolemPII-v2-base 277M 0.886 / 0.951 / 0.917 0.872 / 0.983 / 0.924
GolemPII-v2-tiny 45M 0.844 / 0.927 / 0.884 0.797 / 0.968 / 0.874

By category (entity level, P / R / F1)

Category GolemPII-v2-base GolemPII-v2-tiny
Person 0.888 / 0.986 / 0.934 0.846 / 0.971 / 0.904
Address 0.777 / 1.000 / 0.874 0.735 / 0.985 / 0.842
Phone 0.940 / 0.993 / 0.966 0.883 / 0.965 / 0.922
Email 0.940 / 0.998 / 0.968 0.734 / 1.000 / 0.846
Account (identity, card and bank numbers) 0.762 / 0.905 / 0.827 0.571 / 0.890 / 0.696

By label (character F1)

EMAIL 0.936, PHONE_NUM 0.934, LAST_NAME 0.903, FIRST_NAME 0.898, PASSPORT_NUM 0.835, ID_NUM 0.832, CITY 0.705, CC_NUM 0.539, DATE 0.520, DRIVER_LICENSE_NUM 0.475.

Person names on real text (NEMO, never trained on)

Model Person P / R / F1
GolemPII-v1 0.541 / 0.824 / 0.653
GolemPII-v2-base 0.853 / 0.925 / 0.887
GolemPII-v2-tiny 0.856 / 0.945 / 0.899

Names the model has never seen

FIRST_NAME F1 0.972, LAST_NAME F1 0.959.

Speed and size

Measured on an Apple M4 Pro (48 GB, macOS 27.0.1, torch 2.15.0, transformers 5.22.1) over 1,000 synthetic enterprise-style texts with a mean length of 1,512 characters, with the pipeline settings shown under Usage and batch size 8. Median of three runs; other programs were running, so treat differences of a few percent as noise.

Model Parameters File size Seconds per 1,000 texts, Apple GPU (MPS) Seconds per 1,000 texts, CPU
GolemPII-v2-base 277M 1,110 MB 63.8 115.6
GolemPII-v2-small 125M 502 MB 65.8 120.9
GolemPII-v2-tiny 45M 179 MB 13.0 15.6

The small model is less than half the size of the base model on disk and in memory, and runs at the same speed: both have a 12-layer encoder, and the base model's extra parameters are its larger vocabulary table. The tiny model is about five times faster on the GPU and seven times faster on the CPU.

Intended use and out of scope

Intended use: finding personal information in Hebrew text, and in Hebrew mixed with English, to assist redaction, masking and data-loss-prevention review.

Out of scope or not evaluated:

  • Other languages. Arabic, and English text on its own, were not evaluated.
  • Other sources. Speech transcripts, OCR output and handwriting were not evaluated.
  • Anonymisation. The model is a detector that assists redaction, not a guarantee of anonymisation or of compliance with any privacy law. Text that must be anonymised needs human review.
  • Other entity types. Organisations, job titles, countries and reference numbers (order, ticket and employee numbers) are not labelled.

Limits

All limits of the main model apply. In addition:

  • Precision on account numbers and emails is clearly lower than the main model's: it tags more reference numbers as identity or card numbers, and more non-email strings as emails.
  • One training run; the spread between runs was not measured for this variant.

Training

Same data and settings as the main model, except 10 epochs in place of 3.

Citation

If you use this model in a product, a paper or a derived model, please credit it as: GolemPII-v2-tiny by Liran Baba, with a link to https://huggingface.co/CordwainerSmith/GolemPII-v2-tiny.

@misc{baba2026golempii_v2_tiny,
  author       = {Baba, Liran},
  title        = {GolemPII v2 tiny: personal information detection for Hebrew text},
  year         = {2026},
  version      = {2.0},
  publisher    = {Hugging Face},
  howpublished = {\url{https://huggingface.co/CordwainerSmith/GolemPII-v2-tiny}}
}

Licence and credits

CC-BY-4.0, following the encoder.

  • Encoder: dicta-il/dictabert-tiny by Dicta (CC-BY-4.0).
  • Real training sentences: IAHLT Hebrew named-entity dataset, github.com/IAHLT/hebrew_named_entities_open_dataset (Apache-2.0), with person names replaced.
Downloads last month
48
Safetensors
Model size
44.7M params
Tensor type
F32
Β·
Inference Providers NEW
This model isn't deployed by any Inference Provider. πŸ™‹ Ask for provider support

Model tree for CordwainerSmith/GolemPII-v2-tiny

Finetuned
(1)
this model

Space using CordwainerSmith/GolemPII-v2-tiny 1

Evaluation results