CSVW-EO

CSVW-EO extends the W3C CSV on the Web (CSVW) standard with privacy-safe metadata for:

• Differential Privacy contribution modeling
• Structural dataset assumptions
• Dummy data generation
• Public partition/key definitions
• Validation workflows

Documentation: https://dscc-admin-ch.github.io/csvw-eo-docs/

---

Why CSVW-EO?

Many datasets cannot be shared directly due to privacy or governance constraints.

CSVW-EO allows publishing safe assumptions about datasets without sharing the underlying sensitive data.

Examples include:

• schema information
• nullable proportions
• public partitions
• contribution bounds for DP
• logical dependencies

---

Repository Structure

Component	Description
csvw-eo-library/	Python library
docs/	MkDocs documentation
csvw-eo-vocab.ttl	RDF vocabulary
csvw-eo-constraints.ttl	SHACL validation rules

---

Installation

pip install csvw-eo

Quick Example

from csvw_eo.make_metadata_from_data import make_metadata_from_data
from csvw_eo.make_dummy_from_metadata import make_dummy_from_metadata

metadata = make_metadata_from_data(
    df,
    privacy_unit="user_id",
)
dummy_df = make_dummy_from_metadata(metadata)

Documentation

Section	Link
Vocabulary	https://dscc-admin-ch.github.io/csvw-eo-docs/vocabulary/overview/
Python Library	https://dscc-admin-ch.github.io/csvw-eo-docs/library/overview/
API Reference	https://dscc-admin-ch.github.io/csvw-eo-docs/api/