Update README.md
Browse files
README.md
CHANGED
|
@@ -6,7 +6,15 @@ license: mit
|
|
| 6 |
|
| 7 |
> Merged and simplified dialog act datasets from the [silicone collection](https://huggingface.co/datasets/silicone/)
|
| 8 |
|
| 9 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 10 |
|
| 11 |
```python
|
| 12 |
[
|
|
@@ -24,7 +32,9 @@ license: mit
|
|
| 24 |
]
|
| 25 |
```
|
| 26 |
|
| 27 |
-
|
|
|
|
|
|
|
| 28 |
|
| 29 |
```python
|
| 30 |
mapping = {
|
|
|
|
| 6 |
|
| 7 |
> Merged and simplified dialog act datasets from the [silicone collection](https://huggingface.co/datasets/silicone/)
|
| 8 |
|
| 9 |
+
**Note**: This dataset is highly imbalanced and it is recommended to use a library like [imbalanced-learn](https://imbalanced-learn.org/stable/) before proceeding with training.
|
| 10 |
+
|
| 11 |
+
## Feature description
|
| 12 |
+
|
| 13 |
+
- `text_a`: The utterance prior to the utterance being classified. (Say for dialog with turns 1-2-3, if we are trying to find the dialog act for 2, text_a is 1)
|
| 14 |
+
- `text_b`: The utterance to be classified
|
| 15 |
+
- `label`: Dialog act label (as integer between 0-10, as mapped below)
|
| 16 |
+
|
| 17 |
+
## Labels map
|
| 18 |
|
| 19 |
```python
|
| 20 |
[
|
|
|
|
| 32 |
]
|
| 33 |
```
|
| 34 |
|
| 35 |
+
*****
|
| 36 |
+
|
| 37 |
+
### How the original datasets were mapped:
|
| 38 |
|
| 39 |
```python
|
| 40 |
mapping = {
|