Publish corrected FLEURS test split
Browse files- README.md +61 -0
- data/test-00000-of-00001.parquet +3 -0
README.md
ADDED
|
@@ -0,0 +1,61 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
language:
|
| 3 |
+
- hy
|
| 4 |
+
license: cc-by-4.0
|
| 5 |
+
tags:
|
| 6 |
+
- automatic-speech-recognition
|
| 7 |
+
- armenian
|
| 8 |
+
- fleurs
|
| 9 |
+
pretty_name: FLEURS Armenian Reference Corrections
|
| 10 |
+
size_categories:
|
| 11 |
+
- n<1K
|
| 12 |
+
dataset_info:
|
| 13 |
+
features:
|
| 14 |
+
- name: recording_id
|
| 15 |
+
dtype: string
|
| 16 |
+
- name: file_name
|
| 17 |
+
dtype: string
|
| 18 |
+
- name: fleurs_id
|
| 19 |
+
dtype: int64
|
| 20 |
+
- name: original_transcription
|
| 21 |
+
dtype: string
|
| 22 |
+
- name: corrected_transcription
|
| 23 |
+
dtype: string
|
| 24 |
+
splits:
|
| 25 |
+
- name: test
|
| 26 |
+
num_examples: 314
|
| 27 |
+
configs:
|
| 28 |
+
- config_name: default
|
| 29 |
+
data_files:
|
| 30 |
+
- split: test
|
| 31 |
+
path: data/test-*.parquet
|
| 32 |
+
---
|
| 33 |
+
|
| 34 |
+
# FLEURS Armenian Reference Corrections
|
| 35 |
+
|
| 36 |
+
This private dataset contains manual reference corrections used by
|
| 37 |
+
[ArmBench ASR](https://huggingface.co/spaces/Metric-AI/ArmBench-ASR) for the
|
| 38 |
+
Armenian (`hy_am`) test split of [`google/fleurs`](https://huggingface.co/datasets/google/fleurs).
|
| 39 |
+
|
| 40 |
+
Only rows whose reviewed transcription differs from the original FLEURS
|
| 41 |
+
`raw_transcription` are included. Audio is not duplicated; use `recording_id`
|
| 42 |
+
or `file_name` to join these rows to the upstream FLEURS test split.
|
| 43 |
+
|
| 44 |
+
## Fields
|
| 45 |
+
|
| 46 |
+
- `recording_id`: stable recording identifier, equal to the WAV filename stem.
|
| 47 |
+
- `file_name`: upstream FLEURS WAV filename.
|
| 48 |
+
- `fleurs_id`: original FLEURS sentence ID. This is not unique per recording.
|
| 49 |
+
- `original_transcription`: original upstream `raw_transcription`.
|
| 50 |
+
- `corrected_transcription`: manually reviewed reference used by ArmBench ASR.
|
| 51 |
+
|
| 52 |
+
## Scope
|
| 53 |
+
|
| 54 |
+
- Upstream dataset: `google/fleurs`
|
| 55 |
+
- Configuration: `hy_am`
|
| 56 |
+
- Split compared: `test`
|
| 57 |
+
- Changed references: 314
|
| 58 |
+
- Unchanged reviewed entries excluded: 78
|
| 59 |
+
|
| 60 |
+
Corrections were checked by exact comparison with the upstream transcription.
|
| 61 |
+
This repository is a correction manifest, not a replacement copy of FLEURS.
|
data/test-00000-of-00001.parquet
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:65ec8e9127e0a973f41f18436663e22c73199b0f66721afa9950f37f994f4a3e
|
| 3 |
+
size 87283
|