# Crowdsourced UK and Ireland English Dialect Speech (OpenSLR 83)

## Facts

- Provider: Isin Demirsahin, Oddur Kjartansson, Alexander Gutkin, Clara Rivera (Google)
- License: CC-BY-SA-4.0 (https://www.openslr.org/83/)
- Size: 30 files
- Format: Parquet (in sampled files)
- Stored at: Hugging Face, https://huggingface.co/datasets/ylacombe/english_dialects
- Access: public, readable without an account
- Category: Speech & Audio
- Source: https://aclanthology.org/2020.lrec-1.804/
- Page: https://scigantic.com/archive/97bd4649-a15f-5bfc-8e7a-96e9ab0c1b74
- Keywords: Speech & Audio, speech recognition, text-to-speech, Parquet, Hugging Face

## About

This dataset consists of 31 hours of transcribed high-quality audio of English sentences recorded by 120 volunteers speaking with different accents of the British Isles: Southern England, Midlands, Northern England, Welsh, Scottish and Irish varieties, split by self-identified accent and gender. The recording scripts were written for accent elicitation, covering a range of phonological phenomena with high phoneme coverage, pronunciations of global locations, airlines and common personal names, and native pronunciations of local words; some lines overlap with the CSTR VCTK corpus and the Speech Accent Archive for comparison across resources. It supports accent and dialect identification, accented speech recognition, multi-speaker text-to-speech and sociophonetic analysis. This repository restructures the OpenSLR SLR83 archives for streaming: eleven folders (irish_male, midlands_female, midlands_male, northern_female, northern_male, scottish_female, scottish_male, southern_female, southern_male, welsh_female, welsh_male), each a Parquet file with audio, text and speaker id (measured 2026-11-14; the Hub reports about 60 GB). Load a dialect with datasets.load_dataset('ylacombe/english_dialects', 'welsh_female'). Released by Google on OpenSLR under CC BY-SA 4.0; cite Demirsahin et al., LREC 2020.

## Load the data

```python
from datasets import load_dataset
ds = load_dataset("ylacombe/english_dialects")
print(ds)
```

## Data structure

File types in sampled files: Parquet

Example file: `irish_male/train-00000-of-00001-876ed4aebc6599d3.parquet`

## Open it

- In a hosted notebook with the data mounted: https://scigantic.com/archive/97bd4649-a15f-5bfc-8e7a-96e9ab0c1b74
- Full record as JSON (schema card included): https://api.scigantic.com/api/archives/97bd4649-a15f-5bfc-8e7a-96e9ab0c1b74
- Over MCP: `get_schema_card` with id `97bd4649-a15f-5bfc-8e7a-96e9ab0c1b74` at https://api.scigantic.com/mcp

## Related topics

- [Parquet datasets](https://scigantic.com/topics/parquet)
- [Speech & Audio datasets](https://scigantic.com/topics/speech-and-audio)
- [Speech recognition datasets](https://scigantic.com/topics/speech-recognition)
- [Text-to-speech datasets](https://scigantic.com/topics/text-to-speech)
