Earnings-22: accented earnings call ASR benchmark
- Provider
- Miguel Del Rio, Peter Ha, Quinten McNamara, Corey Miller, Shipra Chandra (Rev.com); Hugging Face conversion by distil-whisper
- License
- CC-BY-SA-4.0 (source)
- Size
- 44 files
- Format
- Parquet (in sampled files)
- Stored at
- Hugging Face (https://huggingface.co/datasets/distil-whisper/earnings22)
- Access
- Free and public. No account with the data provider is needed.
- Category
- Speech & Audio
- Source
- https://arxiv.org/abs/2203.15591
About Earnings-22: accented earnings call ASR benchmark
Earnings-22 is a free-to-use benchmark of real-world, accented English audio released by Rev to bridge academic and industrial speech recognition research. It contains 125 files totalling roughly 119 hours of English-language earnings calls from global companies, with speakers from seven language regions and 27 countries, and provides full audio, transcripts and metadata such as ticker symbol, headquarters country and Rev's 'Language Region'. It is intended to evaluate ASR models on long recordings from several minutes to tens of minutes, reported as word error rate averaged over the 125 files. This Hugging Face conversion from the distil-whisper project has two configurations: full (4 Parquet shards, one row per call) and chunked (38 shards of segmented audio with aligned text) (measured 2026-11-14; the Hub reports about 62 GB). Load with datasets.load_dataset('distil-whisper/earnings22', 'chunked'). The upstream Rev repository licenses Earnings-22 under CC BY-SA 4.0, which governs this copy; cite Del Rio et al., 'Earnings-22: A Practical Benchmark for Accents in the Wild', 2022.
How to load Earnings-22: accented earnings call ASR benchmark in Python
from datasets import load_dataset
ds = load_dataset("distil-whisper/earnings22")
print(ds)Data structure
File types in sampled files: Parquet
Files
Example file: chunked/test-00000-of-00038-6f4f182bdbb6f186.parquet
Analyze this dataset without downloading it
On Scigantic, Earnings-22: accented earnings call ASR benchmark opens in a hosted JupyterLab notebook with the data already mounted, so you can read files where they are stored instead of copying them first. Each notebook includes an AI assistant that can see the mounted files and write and run analysis code. Open the dataset page to launch a notebook, or browse the full catalog.
This page is also available as Markdown and as JSON.
Related Speech & Audio datasets
- Hifitts 2
- RixVox-v2: Swedish parliamentary speech dataset
- Hi-Fi Multi-Speaker English TTS Dataset (Hi-Fi TTS)
- LJ Speech single-speaker English TTS dataset
- YODAS2: YouTube-Oriented Dataset for Audio and Speech (long-form)
- LibriTTS-R: restored LibriTTS speech corpus
- LibriTTS multi-speaker English text-to-speech corpus
- LibriSpeech-Long: Long-Form Speech Benchmark of Contiguous Audiobook Segments up to 4 Minutes (Google DeepMind)