Datasets:
Tasks:
Text Generation
Sub-tasks:
language-modeling
Languages:
German
Size:
10M - 100M
ArXiv:
License:
Update DATASHEET.md
Browse files- DATASHEET.md +3 -1
DATASHEET.md
CHANGED
|
@@ -55,7 +55,9 @@ GPT-2 tokens.
|
|
| 55 |
Each instance includes: a unique identifier for source
|
| 56 |
cross-referencing, source dataset name, quality-filtered and
|
| 57 |
paragraph-deduplicated raw text, canonical SPDX license URL, thematic
|
| 58 |
-
domain key,
|
|
|
|
|
|
|
| 59 |
|
| 60 |
### Is there a label or target associated with each instance?
|
| 61 |
|
|
|
|
| 55 |
Each instance includes: a unique identifier for source
|
| 56 |
cross-referencing, source dataset name, quality-filtered and
|
| 57 |
paragraph-deduplicated raw text, canonical SPDX license URL, thematic
|
| 58 |
+
domain key, GPT-2 token count, a perplexity score calculated using a
|
| 59 |
+
KenLM model trained on German Wikipedia text, and a OCR quality score
|
| 60 |
+
calculated using [OCRoscope](https://github.com/Pleias/OCRoscope).
|
| 61 |
|
| 62 |
### Is there a label or target associated with each instance?
|
| 63 |
|