Create README.md
Browse files
README.md
ADDED
@@ -0,0 +1,133 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
---
|
2 |
+
license: apache-2.0
|
3 |
+
dataset_info:
|
4 |
+
features:
|
5 |
+
- name: hypothesis
|
6 |
+
sequence: string
|
7 |
+
- name: transcription
|
8 |
+
dtype: string
|
9 |
+
- name: input1
|
10 |
+
dtype: string
|
11 |
+
- name: hypothesis_concatenated
|
12 |
+
dtype: string
|
13 |
+
- name: source
|
14 |
+
dtype: string
|
15 |
+
- name: id
|
16 |
+
dtype: string
|
17 |
+
- name: dummy_str
|
18 |
+
dtype: string
|
19 |
+
- name: dummy_list
|
20 |
+
sequence: 'null'
|
21 |
+
- name: prompt
|
22 |
+
dtype: string
|
23 |
+
splits:
|
24 |
+
- name: train
|
25 |
+
num_bytes: 469086507
|
26 |
+
num_examples: 286366
|
27 |
+
- name: test
|
28 |
+
num_bytes: 24103011
|
29 |
+
num_examples: 18237
|
30 |
+
download_size: 125101353
|
31 |
+
dataset_size: 493189518
|
32 |
+
configs:
|
33 |
+
- config_name: default
|
34 |
+
data_files:
|
35 |
+
- split: train
|
36 |
+
path: data/train-*
|
37 |
+
- split: test
|
38 |
+
path: data/test-*
|
39 |
+
---
|
40 |
+
|
41 |
+
# Dataset Name: Pilot dataset for Multi-domain ASR corrections
|
42 |
+
|
43 |
+
<p align="center"> <img src="hypilot.jpg" height ="100"> </p>
|
44 |
+
|
45 |
+
|
46 |
+
## Description
|
47 |
+
|
48 |
+
This dataset is a pilot version of a larger dataset for automatic speech recognition (ASR) corrections across multiple domains.
|
49 |
+
It contains paired hypotheses and corrected transcriptions for various ASR tasks consolidated from [PeacefulData/HyPoradise-v0](https://huggingface.co/datasets/PeacefulData/HyPoradise-v0)
|
50 |
+
|
51 |
+
## Structure
|
52 |
+
|
53 |
+
### Data Split
|
54 |
+
|
55 |
+
The dataset is divided into training and test splits:
|
56 |
+
|
57 |
+
- Training Data: 281,082 entries
|
58 |
+
- Approximately 6,255,198 tokens for transcriptions
|
59 |
+
- Approximately 31,211,083 tokens for concatenated hypotheses
|
60 |
+
- Test Data: 16,108 entries
|
61 |
+
- Approximately 327,750 tokens for transcriptions
|
62 |
+
- Approximately 1,629,093 tokens for concatenated hypotheses
|
63 |
+
|
64 |
+
### Columns
|
65 |
+
|
66 |
+
- `hypothesis`: N-best hypothesis from beam search.
|
67 |
+
- `transcription`: Corrected asr transcription.
|
68 |
+
- `hypothesis_concatenated`: An alternative version of the text output.
|
69 |
+
- `source`: The source of the text entry, indicating the origin dataset.
|
70 |
+
- `prompt`: Instructional prompt for correction task
|
71 |
+
- `score`: An acoustic model score (not all entries have this).
|
72 |
+
|
73 |
+
### Source Datasets
|
74 |
+
|
75 |
+
The dataset combines entries from various sources:
|
76 |
+
|
77 |
+
- **Training Sources**:
|
78 |
+
- `train_td3`: 50,000 entries
|
79 |
+
- `train_other_500`: 50,000 entries
|
80 |
+
- `train_cv`: 47,293 entries
|
81 |
+
- `train_lrs2`: 42,940 entries
|
82 |
+
- `train_wsj_score`: 37,514 entries
|
83 |
+
- `train_swbd`: 36,539 entries
|
84 |
+
- `train_chime4`: 9,600 entries
|
85 |
+
- `train_atis`: 3,964 entries
|
86 |
+
- `train_coraal`: 3,232 entries
|
87 |
+
- **Test Sources**:
|
88 |
+
- `test_ls_other`: 2,939 entries
|
89 |
+
- `test_ls_clean`: 2,620 entries
|
90 |
+
- `test_lrs2`: 2,259 entries
|
91 |
+
- `test_swbd`: 2,000 entries
|
92 |
+
- `test_cv`: 2,000 entries
|
93 |
+
- `test_chime4`: 1,320 entries
|
94 |
+
- `test_td3`: 1,155 entries
|
95 |
+
- `test_wsj_score`: 836 entries
|
96 |
+
- `test_atis`: 809 entries
|
97 |
+
- `test_coraal`: 170 entries
|
98 |
+
|
99 |
+
|
100 |
+
## Access
|
101 |
+
|
102 |
+
The dataset can be accessed and downloaded through the HuggingFace Datasets library. Use the following command to load the dataset:
|
103 |
+
|
104 |
+
```python
|
105 |
+
from datasets import load_dataset
|
106 |
+
dataset = load_dataset("PeacefulData/HyPoradise-pilot")
|
107 |
+
```
|
108 |
+
|
109 |
+
## Acknowledgments
|
110 |
+
|
111 |
+
This dataset is consolidated from the PeacefulData/HyPoradise-v0 dataset. Thanks to the original creators for making this data available.
|
112 |
+
|
113 |
+
### References
|
114 |
+
|
115 |
+
```bib
|
116 |
+
@inproceedings{yang2023generative,
|
117 |
+
title={Generative speech recognition error correction with large language models and task-activating prompting},
|
118 |
+
author={Yang, Chao-Han Huck and Gu, Yile and Liu, Yi-Chieh and Ghosh, Shalini and Bulyko, Ivan and Stolcke, Andreas},
|
119 |
+
booktitle={2023 IEEE Automatic Speech Recognition and Understanding Workshop (ASRU)},
|
120 |
+
pages={1--8},
|
121 |
+
year={2023},
|
122 |
+
organization={IEEE}
|
123 |
+
}
|
124 |
+
```
|
125 |
+
|
126 |
+
```bib
|
127 |
+
@inproceedings{chen2023hyporadise,
|
128 |
+
title={HyPoradise: An Open Baseline for Generative Speech Recognition with Large Language Models},
|
129 |
+
author={CHEN, CHEN and Hu, Yuchen and Yang, Chao-Han Huck and Siniscalchi, Sabato Marco and Chen, Pin-Yu and Chng, Ensiong},
|
130 |
+
booktitle={Thirty-seventh Conference on Neural Information Processing Systems Datasets and Benchmarks Track},
|
131 |
+
year={2023}
|
132 |
+
}
|
133 |
+
```
|