---
datasets:
- vuihocrnd/l2-arctic-cleaned
language:
- en
license: mit
task_categories:
- audio-classification
pretty_name: L2 Arctic Cleaned
size_categories:
- 1K<n<10K
---

# L2 Arctic Cleaned

Cleaned version of the **L2 Arctic** dataset.  
It includes English L2 speech recordings split into **train** and **test**, with aligned metadata CSVs.
Test data contains the recording data of two Vietnamese people 
(HQTV-Male, PNV-Female)

## Dataset Structure

- `train/`
  - `train/audio/` — train audio files (`.wav`)  
  - `train/trainset.csv` — train data with columns: `Label,Canonical,Transcript,Error`
  - `train/metadata.csv` — metadata with columns: `file_name,Label`
    
- `test/`
  - `test/audio/` — test audio files (`.wav`)
  - `test/trainset.csv` — train data with columns: `Label,Canonical,Transcript,Error`
  - `test/metadata.csv` — metadata with columns: `file_name,Label`