Datasets:
Commit ·
10afaf4
1
Parent(s): 9a56cf6
Update README.md
Browse files
README.md
CHANGED
|
@@ -7,7 +7,7 @@ dataset_info:
|
|
| 7 |
dtype: string
|
| 8 |
- name: text
|
| 9 |
dtype: string
|
| 10 |
-
- name:
|
| 11 |
dtype:
|
| 12 |
class_label:
|
| 13 |
names:
|
|
@@ -17,6 +17,17 @@ dataset_info:
|
|
| 17 |
'3': anger
|
| 18 |
'4': fear
|
| 19 |
'5': surprise
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 20 |
splits:
|
| 21 |
- name: train
|
| 22 |
num_bytes: 72761213.15815398
|
|
@@ -46,13 +57,10 @@ language:
|
|
| 46 |
size_categories:
|
| 47 |
- 100K<n<1M
|
| 48 |
---
|
| 49 |
-
|
| 50 |
|
| 51 |
-
The dataset was machine translated
|
| 52 |
|
| 53 |
-
Tool used for translation: [deep-translator](https://pypi.org/project/deep-translator/)
|
| 54 |
-
|
| 55 |
-
Translation script:
|
| 56 |
```python
|
| 57 |
import pandas as pd
|
| 58 |
from deep_translator import GoogleTranslator
|
|
@@ -92,7 +100,7 @@ batch_size = 500
|
|
| 92 |
translated_dataset = df.groupby(df.index // batch_size, group_keys=False).apply(translate_samples)
|
| 93 |
```
|
| 94 |
|
| 95 |
-
|
| 96 |
```yaml
|
| 97 |
0: sadness
|
| 98 |
1: joy
|
|
@@ -100,4 +108,27 @@ Dataset contains 6 emotion labels. Labels are as follows:
|
|
| 100 |
3: anger
|
| 101 |
4: fear
|
| 102 |
5: surprise
|
| 103 |
-
```
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 7 |
dtype: string
|
| 8 |
- name: text
|
| 9 |
dtype: string
|
| 10 |
+
- name: labels
|
| 11 |
dtype:
|
| 12 |
class_label:
|
| 13 |
names:
|
|
|
|
| 17 |
'3': anger
|
| 18 |
'4': fear
|
| 19 |
'5': surprise
|
| 20 |
+
- name: labels_ekman
|
| 21 |
+
dtype:
|
| 22 |
+
class_label:
|
| 23 |
+
names:
|
| 24 |
+
'0': anger
|
| 25 |
+
'1': disgust
|
| 26 |
+
'2': fear
|
| 27 |
+
'3': joy
|
| 28 |
+
'4': sadness
|
| 29 |
+
'5': surprise
|
| 30 |
+
'6': neutral
|
| 31 |
splits:
|
| 32 |
- name: train
|
| 33 |
num_bytes: 72761213.15815398
|
|
|
|
| 57 |
size_categories:
|
| 58 |
- 100K<n<1M
|
| 59 |
---
|
| 60 |
+
The original dataset: https://www.kaggle.com/datasets/nelgiriyewithana/emotions
|
| 61 |
|
| 62 |
+
The derived dataset was machine translated from English into Latvian using the free Google Translate API (with [deep-translator](https://pypi.org/project/deep-translator/)). The translation script:
|
| 63 |
|
|
|
|
|
|
|
|
|
|
| 64 |
```python
|
| 65 |
import pandas as pd
|
| 66 |
from deep_translator import GoogleTranslator
|
|
|
|
| 100 |
translated_dataset = df.groupby(df.index // batch_size, group_keys=False).apply(translate_samples)
|
| 101 |
```
|
| 102 |
|
| 103 |
+
Column `labels` contain the following classes:
|
| 104 |
```yaml
|
| 105 |
0: sadness
|
| 106 |
1: joy
|
|
|
|
| 108 |
3: anger
|
| 109 |
4: fear
|
| 110 |
5: surprise
|
| 111 |
+
```
|
| 112 |
+
|
| 113 |
+
Column `labels_ekman` contains the Ekman emotion classes:
|
| 114 |
+
```yaml
|
| 115 |
+
0: anger
|
| 116 |
+
1: disgust - omitted in this dataset
|
| 117 |
+
2: fear
|
| 118 |
+
3: joy
|
| 119 |
+
4: sadness
|
| 120 |
+
5: surprise
|
| 121 |
+
6: neutral - omitted in this dataset
|
| 122 |
+
```
|
| 123 |
+
which were mapped from the original classes as follows:
|
| 124 |
+
```yaml
|
| 125 |
+
Original -> Ekman
|
| 126 |
+
sadness (0) -> sadness (4)
|
| 127 |
+
joy (1) -> joy (3)
|
| 128 |
+
love (2) -> joy (3)
|
| 129 |
+
anger (3) -> anger (0)
|
| 130 |
+
fear (4) -> fear (2)
|
| 131 |
+
surprise (5) -> surprise (5)
|
| 132 |
+
```
|
| 133 |
+
|
| 134 |
+
|