Add "Bias, Risks, and Limitations" section to model card
#15
by
giskard-bot
- opened
README.md
CHANGED
|
@@ -130,3 +130,13 @@ Output:
|
|
| 130 |
}
|
| 131 |
|
| 132 |
```
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 130 |
}
|
| 131 |
|
| 132 |
```
|
| 133 |
+
|
| 134 |
+
|
| 135 |
+
## Bias, Risks, and Limitations
|
| 136 |
+
|
| 137 |
+
The model was analyzed with [Giskard](github.com/giskard-AI/giskard) and the following limitations were identified:
|
| 138 |
+
|
| 139 |
+
- Typos and misspellings can have a negative impact on the model performances, potentially affecting non-native speakers or people with limited educational background.
|
| 140 |
+
- The model is sensitive to letter case and could classify the same text differently if transformed to lowercase or uppercase. Please make sure that your usage takes this into account.
|
| 141 |
+
|
| 142 |
+
A full report is available [here](https://htmlpreview.github.io/?https://github.com/Giskard-AI/ml-scan-reports/blob/main/scan_reports/huggingface/cardiffnlp/twitter-roberta-base-sentiment-latest/ea9d76e8-a9f2-429a-b053-ed6da8ef6df/report.html).
|