Update README.md
Browse files
README.md
CHANGED
@@ -62,6 +62,59 @@ configs:
|
|
62 |
---
|
63 |
# speechocean762: A non-native English corpus for pronunciation scoring task
|
64 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
65 |
## Introduction
|
66 |
Pronunciation scoring is a crucial technology in computer-assisted language learning (CALL) systems. The pronunciation quality scores might be given at phoneme-level, word-level, and sentence-level for a typical pronunciation scoring task.
|
67 |
|
|
|
62 |
---
|
63 |
# speechocean762: A non-native English corpus for pronunciation scoring task
|
64 |
|
65 |
+
## How to use?
|
66 |
+
|
67 |
+
you can load data using
|
68 |
+
|
69 |
+
```py
|
70 |
+
speechocean762_dataset = load_dataset('seba3y/speechocean762')
|
71 |
+
```
|
72 |
+
```py
|
73 |
+
>> speechocean762_dataset
|
74 |
+
DatasetDict({
|
75 |
+
train: Dataset({
|
76 |
+
features: ['spk', 'age', 'gender', 'utt_name', 'audio', 'utt_text', 'utt_accuracy', 'utt_completeness', 'utt_fluency', 'utt_prosodic', 'utt_total', 'words', 'words_accuracy', 'words_stress', 'words_total', 'phones', 'phones_godness'],
|
77 |
+
num_rows: 2500
|
78 |
+
})
|
79 |
+
test: Dataset({
|
80 |
+
features: ['spk', 'age', 'gender', 'utt_name', 'audio', 'utt_text', 'utt_accuracy', 'utt_completeness', 'utt_fluency', 'utt_prosodic', 'utt_total', 'words', 'words_accuracy', 'words_stress', 'words_total', 'phones', 'phones_godness'],
|
81 |
+
num_rows: 2500
|
82 |
+
})
|
83 |
+
})
|
84 |
+
```
|
85 |
+
Features are ordered as following
|
86 |
+
1- Demographic featurs: `'spk', 'age', 'gender', 'utt_name'`
|
87 |
+
2- Sentence-level featurs: `'audio', 'utt_text', 'utt_accuracy', 'utt_completeness', 'utt_fluency', 'utt_prosodic', 'utt_total'`
|
88 |
+
3- Word-level featurs: `'words', 'words_accuracy', 'words_stress', 'words_total'`
|
89 |
+
4- Phoneme-level featurs: `'phones', 'phones_godness'`
|
90 |
+
|
91 |
+
```py
|
92 |
+
>> speechocean762_dataset['train'][0]
|
93 |
+
```
|
94 |
+
```py
|
95 |
+
{'spk': 1,
|
96 |
+
'age': 6,
|
97 |
+
'gender': 'm',
|
98 |
+
'utt_name': 10011,
|
99 |
+
'audio': {'path': '000010011.WAV',
|
100 |
+
'array': array([-9.46044922e-04, -2.38037109e-03, -1.31225586e-03, ...,
|
101 |
+
-9.15527344e-05, 3.05175781e-04, -2.44140625e-04]),
|
102 |
+
'sampling_rate': 16000},
|
103 |
+
'utt_text': 'WE CALL IT BEAR',
|
104 |
+
'utt_accuracy': 8,
|
105 |
+
'utt_completeness': 10.0,
|
106 |
+
'utt_fluency': 9,
|
107 |
+
'utt_prosodic': 9,
|
108 |
+
'utt_total': 8,
|
109 |
+
'words': "['WE', 'CALL', 'IT', 'BEAR']",
|
110 |
+
'words_accuracy': '[10, 10, 10, 6]',
|
111 |
+
'words_stress': '[10, 10, 10, 10]',
|
112 |
+
'words_total': '[10, 10, 10, 6]',
|
113 |
+
'phones': "[['W', 'IY0'], ['K', 'AO0', 'L'], ['IH0', 'T'], ['B', 'EH0', 'R']]",
|
114 |
+
'phones_godness': '[[2.0, 2.0], [2.0, 1.8, 1.8], [2.0, 2.0], [2.0, 1.0, 1.0]]'}
|
115 |
+
```
|
116 |
+
For word-level features, the 'words' in each sample is a list containing words, while 'words_accuracy', 'words_stress', and 'words_total' are lists of the same length as the words. The mapping is such that the first word corresponds to the first value in 'words_accuracy', and so on. On the other hand, for phoneme-level features, the 'phones' in each sample is a 2D list, with each sublist corresponding to a single word
|
117 |
+
|
118 |
## Introduction
|
119 |
Pronunciation scoring is a crucial technology in computer-assisted language learning (CALL) systems. The pronunciation quality scores might be given at phoneme-level, word-level, and sentence-level for a typical pronunciation scoring task.
|
120 |
|