README.md

metadata

license: cc-by-4.0
pretty_name: Swimming Resources Database
language:
  - en
task_categories:
  - text-retrieval
  - question-answering
  - text-classification
tags:
  - swimming
  - sports
  - open-data
  - coaching
  - water-safety
size_categories:
  - n<1K
configs:
  - config_name: default
    data_files:
      - split: train
        path: swimming_resources.parquet

Swimming Resources Database

The Swimming Resources Database is an open dataset of 410 verified swimming resources across 20 categories, covering governance, learn-to-swim programs, technique, training, apps, books, coaching education, sports science, news, gear, open water, triathlon, masters swimming, para swimming, drowning prevention, records, college programs, recruiting services, and camps.

The dataset is intended for retrieval-augmented generation, recommendation systems, knowledge bases, directory websites, and research workflows that need structured information about the swimming ecosystem.

Dataset Description

• Repository: timothy22000/swimming-resources
• Point of Contact: @timothy22000
• Resource count: 410 across 20 categories
• Schema fields: 16, validated against the project JSON schema on every commit
• Output formats on GitHub: JSON, JSONL, CSV, Parquet (the Parquet shipped here is the same file)

Background

I started collecting these resources while learning to swim as an adult. The useful material was scattered across forums, blogs, and outdated directories, with frequent rebrands and dead links. This dataset is the index I wish had existed when I started, cleaned up and released so other swimmers, coaches, researchers, and ML systems can build on it instead of re-doing the search.

Dataset Files

This Hugging Face dataset repo contains:

File	Description
swimming_resources.parquet	Primary Hugging Face dataset file.
README.md	This dataset card.

Source Data

The canonical source of truth for this dataset is the GitHub repository:

github.com/timothy22000/swimming-resources

Each entry is hand-curated as a Python literal in scripts/data_part1.py through data_part9.py. The build pipeline at scripts/build_dataset.py validates every record against data/schema.json and emits four formats: pretty-printed JSON, JSONL, CSV, and the Parquet file shipped here. The GitHub repository keeps all four output formats checked in alongside the schema, so the same dataset is available without going through datasets.

Two scheduled workflows in the source repository keep the dataset honest:

• validate.yml re-runs the build and schema validation on every pull request and push to main.
• link-check.yml runs weekly, fetches every URL in the dataset, and opens an issue listing any that no longer resolve.

Corrections, additions, and dead-link reports should go through GitHub (issues or pull requests) rather than the Hugging Face discussions tab, so that fixes flow through the validation pipeline before the Parquet here is updated. The contributor workflow is documented in CONTRIBUTING.md.

Loading

from datasets import load_dataset

dataset = load_dataset("t22000t/swimming-resources")
print(dataset["train"][0])

Schema

Each row represents one resource.

Field	Type	Description
id	string	Stable identifier formatted as <prefix>-NNN or <prefix>-NNNN.
category	string	One of the 20 category identifiers listed below.
subcategory	string	Optional finer grouping within a category.
name	string	Official or commonly used display name.
type	string	Resource type, such as federation, app, book, curriculum, or database.
audience	string	Main audience served by the resource.
description	string	Concise one-sentence summary.
url	string	Canonical HTTP or HTTPS URL.
pricing	string	One of free, paid, freemium, mixed, or varies.
region	string	global, a continent, a country, or another practical region label.
language	string	ISO 639-1 language code, optionally with a region suffix.
year_founded	int64 or null	Known founding year, when available.
status	string	One of active, rebranded, or defunct.
notes	string	Caveats, rebrand details, ISBNs, replacement products, or context.
tags	list[string]	Search and retrieval tags.
verified	string	Month when the URL was last verified, formatted YYYY-MM.

Categories

Category ID	Label
governing_bodies	Governing Bodies & Federations
learn_to_swim	Learn-to-Swim & Water Safety Programs
technique	Technique & Stroke Resources
youtube	YouTube Channels
training_plans	Training Plans & Workout Repositories
apps	Apps & Swim Technology
books	Books on Swimming
certifications	Coaching Certifications & Education
science	Sports Science, Biomechanics & Research
news	News, Communities & Magazines
podcasts	Podcasts
gear	Gear, Equipment & Reviews
open_water	Open Water & Marathon Swimming
triathlon	Triathlon Swimming
masters	Masters Swimming
para	Para Swimming & Adaptive Resources
drowning_prevention	Drowning Prevention & Public Health
records	Records, Rankings & Results Databases
college	College / NCAA Programs & Conferences
recruiting	Recruiting Services & Swim Camps

Data Quality

All entries are validated against the project JSON Schema before release. URLs are stored with a verified month and are checked by scheduled automation in the source repository.

The dataset includes historical rebranded and defunct entries when they are useful for search, citation, or migration from older names and products.

Licensing

The database content is released under Creative Commons Attribution 4.0 International (CC BY 4.0). Build scripts in the source repository are released under the MIT License.

Please attribute the dataset to Timothy Sum and link to the source repository and this Hugging Face dataset page.

Citation

@misc{swimming_resources_db_2026,
  title  = {Swimming Resources Database},
  author = {Timothy Sum},
  year   = {2026},
  url    = {https://github.com/timothy22000/swimming-resources},
  note   = {Open-source index of swimming resources, v1.0.0}
}