Data Hierarchy (Concise Model)

We define three structural object types:

1. Catalog
2. Dataset
3. Data File

---

1. Catalog

A listing of datasets.

Examples:

• Institutional portal (World Bank)
• Thematic grouping (Human Rights Data)
• Your own curated grouping

A collection is special case of a thematic catalog – often small.

So:

Collection = special, small Catalog (semantic subtype)

A catalog:

• Contains datasets
• May contain sub-catalogs
• Does not itself need to be data

---

2. Dataset

A coherent structured data concept defined by schema and coverage.

Important:

• A dataset is a logical unit.
• It may contain multiple data files.
• It may have sub-datasets (handled via parent relationship).

Examples:

• “US GDP 1900–2020”
• “Human Rights Violations Index”

---

3. Data File

A concrete file artifact.

Examples:

• gdp.csv
• index_2024.json

This is the storage layer.

---

Hierarchy Diagram

Catalog
│
├── Dataset
│   ├── Dataset (sub-dataset)
│   │   └── Data File
│   └── Data File
│
└── Dataset
    └── Data File

---

Parent Model

Keep it simple:

type: catalog | dataset | datafile
parent: <id>   # single structural parent

Rules:

• Catalog → contains datasets
• Dataset → contains datasets or datafiles
• Datafile → leaf node

Avoid multi-parent structural trees.

---

Core Principle

Everything is a node in a typed hierarchy.

Catalogs organize. Datasets define structure. Data files implement datasets.

That’s enough structure to scale without overengineering.