Skip to content

Dataset · edinburghcstr

English dialogues with diverse accents

edinburghcstr/edacc

40 hours of paired English conversations with diverse accents for ASR training and evaluation.

What's inside

The corpus contains 40 hours of paired English dialogues between people with various accents (both native and non-native), totaling 10K to 100K records. Each participant includes a profile describing their linguistic background. Audio and text modalities are present in the dataset.

What you can do with it

  • Evaluate ASR model robustness to various accents
  • Fine-tune ASR systems for non-native English
  • Classify audio by language profile or speaker accent
Tasks
Automatic speech recognition, Audio classification
Size
10K–100K rows
Downloads this month
6.6K
Likes
18

Models trained on it

How to load it

from datasets import load_dataset

ds = load_dataset("edinburghcstr/edacc")