Update README.md
Browse files
README.md
CHANGED
|
@@ -1,11 +1,11 @@
|
|
| 1 |
-
---
|
| 2 |
-
license: cc
|
| 3 |
-
task_categories:
|
| 4 |
-
- text-classification
|
| 5 |
-
- feature-extraction
|
| 6 |
-
language:
|
| 7 |
-
- en
|
| 8 |
-
---
|
| 9 |
# Text Quality Assessment Dataset
|
| 10 |
|
| 11 |
## Overview
|
|
@@ -37,18 +37,11 @@ The quality scores of each text were assessed using
|
|
| 37 |
- Yeo-Johnson transformed
|
| 38 |
|
| 39 |
3. **NVIDIA Quality Score**:
|
| 40 |
-
-
|
| 41 |
-
- "Low" (0)
|
| 42 |
-
- "Medium" (0.5)
|
| 43 |
-
- "High" (1)
|
| 44 |
-
- Weighted by predicted probabilities (result between 0 and 1)
|
| 45 |
-
- Logit transformed
|
| 46 |
-
- Yeo-Johnson transformed
|
| 47 |
|
| 48 |
-
|
| 49 |
- First principal component of fineweb-edu and NVIDIA scores
|
| 50 |
-
|
| 51 |
-
All scores adjusted for length using linear regression with the transformed text length
|
| 52 |
|
| 53 |
## Key Insights
|
| 54 |
|
|
|
|
| 1 |
+
---
|
| 2 |
+
license: cc
|
| 3 |
+
task_categories:
|
| 4 |
+
- text-classification
|
| 5 |
+
- feature-extraction
|
| 6 |
+
language:
|
| 7 |
+
- en
|
| 8 |
+
---
|
| 9 |
# Text Quality Assessment Dataset
|
| 10 |
|
| 11 |
## Overview
|
|
|
|
| 37 |
- Yeo-Johnson transformed
|
| 38 |
|
| 39 |
3. **NVIDIA Quality Score**:
|
| 40 |
+
- Logits of "High" quality level - logits of "Low" quality level
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 41 |
|
| 42 |
+
5. **Composite Quality Score**:
|
| 43 |
- First principal component of fineweb-edu and NVIDIA scores
|
| 44 |
+
- Adjusted for length using linear regression with the transformed text length
|
|
|
|
| 45 |
|
| 46 |
## Key Insights
|
| 47 |
|