Practical overview
When it fits
Use it when you need a research workflow for creating PII examples, preparing NER datasets, or measuring Presidio and recognizer behavior. Its notebooks cover generation, exploratory analysis, dataset splitting, and analyzer evaluation.
Who it is for
- Teams and researchers developing or evaluating PII detection models or Presidio recognizers.
- People generating or augmenting named-entity recognition datasets with synthetic examples.
Documented uses
- Generate synthetic sentences by combining templates with fake PII values.
- Tokenize and annotate generated data with entity tags and spans.
- Prepare synthetic data for training, testing, and validation while keeping templates in separate splits.
- Evaluate Presidio Analyzer or custom recognizers and analyze results.
- Represent datasets in a standardized structure and convert them to formats such as CoNLL.
Requirements & limits
The README documents installation from PyPI with pip or from source using uv. It specifies spaCy pipelines for tokenization and the default Presidio configuration.
- The documented default setup uses spaCy en_core_web_sm for tokenization and en_core_web_lg for the default Presidio configuration.
- The README cautions that its vanilla Presidio evaluation example may not produce very accurate results and points to custom analyzer configuration.
- Flair and Stanza are no longer supported dependencies; the README recommends using Presidio Analyzer directly for custom NER models.
- Template-based splitting to avoid leakage between dataset folds is described for synthetically generated data.
Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.
Install Presidio Evaluator from PyPI
Install the package and the listed spaCy models for tokenization and NER.
Before you start
- Python with pip available.
Steps
Install the package with the command below.
pip install presidio-evaluatorDownload the tokenization pipeline with the command below.
python -m spacy download en_core_web_smDownload the NER pipeline with the command below.
python -m spacy download en_core_web_lg
Expected outcome
Presidio Evaluator and the two listed spaCy pipelines are installed.
Guide limitations
- The instructions do not include a post-install verification step for the PyPI installation path.
Install Presidio Evaluator from source
Set up the repository with development dependencies and the spaCy pipelines used by the project.
Before you start
- A way to clone the repository.
- Python with pip available.
Steps
Clone the repository.
Install uv if needed with the command below, then install the package and development dependencies with the command below.
pip install uvuv sync --extra devDownload the tokenization pipeline with the command below.
uv run python -m spacy download en_core_web_smFor the default Presidio configuration, download the NER model with the command below.
uv run python -m spacy download en_core_web_lgRun the documented installation check: the command below.
uv run pytest
Expected outcome
The source installation is set up, and the documented test command can be used to verify it.
Guide limitations
- Flair and Stanza are no longer supported; the README recommends using Presidio Analyzer directly to add custom NER models.
Sources & review
Source record details
- Source availability
- Active
- Sources checked
- 13 October 2026
- Original website title
- GitHub - data-privacy-stack/presidio-research: This package features data-science related tasks for developing new recognizers for Presidio. It is used for the evaluation of the entire system, as well as for evaluating specific PII recogniz · Official homepage title · retained in its source language