shelter-dogs

---
license: mit
task_categories:
  - text-classification
  - sentence-similarity
language:
  - en
tags:
  - dogs
  - adoption
  - shelter
  - petfinder
  - animals
  - pets
size_categories:
  - 10K<n<100K
pretty_name: US Shelter Dogs (PetFinder Snapshot)
---

# 🐾 US Shelter Dogs — PetFinder Adoption Dataset

A cleaned and enriched dataset of **49,036 real adoptable dogs** from US shelters and rescues, originally collected from the PetFinder API on September 20, 2019.

## Dataset Description

Each row represents a single adoptable dog with:
- **Rich text descriptions** written by shelters/rescues (median 708 characters)
- **Breed information** (216 unique breeds, primary + secondary + mixed flag)
- **Demographics**: age class, sex, size, coat type
- **Compatibility flags**: good with children, dogs, cats
- **Training/needs**: house trained, special needs, vaccination status
- **Location**: city and state
- **PetFinder URL**: direct link to the dog's original listing

### `match_text` Column
A pre-built combined text field that merges the dog's structured metadata (name, breed, age, size, compatibility, training status) with their free-text description. Designed for semantic similarity matching with sentence-transformers.

## Source

Originally collected by [Amber Thomas](https://pudding.cool/author/amber-thomas/) for The Pudding's essay ["Finding Forever Homes"](https://pudding.cool/2019/10/shelters/) (October 2019). Raw data from [the-pudding/data](https://github.com/the-pudding/data/tree/master/dog-shelters).

Data was collected via PetFinder's V2 API (`get-animals`) for all US states + DC. Descriptions were obtained via the V1 API. This dataset represents a single-day snapshot of all adoptable dogs on PetFinder.

## Cleaning Applied

1. Removed dogs without descriptions (8,705 rows) and very short descriptions < 30 chars (439 rows)
2. Dropped administrative columns (`declawed`, `stateQ`, `accessed`, `type.x`, `type.y`, `org_id`, `contact_zip`, `contact_country`, `status`, `species`)
3. Created `match_text` combining structured metadata + description for embedding
4. Standardized boolean and compatibility columns

## Column Reference

| Column | Type | Description |
|--------|------|-------------|
| `id` | int | Unique PetFinder ID |
| `url` | string | PetFinder listing URL |
| `name` | string | Dog's name |
| `breed_primary` | string | Primary breed (216 unique) |
| `breed_secondary` | string | Secondary breed (if mixed) |
| `breed_mixed` | bool | Whether mixed breed |
| `age` | string | Baby, Young, Adult, Senior |
| `sex` | string | Male, Female, Unknown |
| `size` | string | Small, Medium, Large, Extra Large |
| `coat` | string | Short, Medium, Long, Wire, Curly, Hairless |
| `env_children` | string | yes / no / unknown |
| `env_dogs` | string | yes / no / unknown |
| `env_cats` | string | yes / no / unknown |
| `house_trained` | bool | House trained status |
| `special_needs` | bool | Has special needs |
| `description` | string | Full shelter-written description |
| `match_text` | string | Combined metadata + description for embedding |
| `contact_city` | string | Shelter city |
| `contact_state` | string | Shelter state |

## Usage

Used by the [Shelter Dog Personality Matcher](https://huggingface.co/spaces/meganariley/shelter-dog-matcher) Gradio Space — a quiz-based app that matches users to adoptable dogs using sentence-transformer semantic similarity.

## Statistics

- **49,036** dogs with descriptions
- **216** unique primary breeds
- **Age**: Adult (57%), Young (33%), Baby (19%), Senior (9%)
- **Size**: Medium (61%), Large (32%), Small (24%), Extra Large (2%)
- **Top breeds**: Pit Bull Terrier, Labrador Retriever, Chihuahua, Mixed Breed, Terrier
- **All 50 US states + DC** represented

## License

The original data was collected from the publicly available PetFinder API. Released under MIT license for research and educational purposes.