Upload README.md with huggingface_hub
Browse files
README.md
CHANGED
|
@@ -15,12 +15,14 @@ The dataset consists of several alignment outputs based on different cosine simi
|
|
| 15 |
|
| 16 |
threshold_matches/
|
| 17 |
|
| 18 |
-
├── matched_{
|
| 19 |
|
| 20 |
-
├──
|
| 21 |
|
| 22 |
...
|
| 23 |
|
|
|
|
|
|
|
| 24 |
|
| 25 |
## Format
|
| 26 |
|
|
@@ -54,9 +56,9 @@ A negative gold standard was used to ensure quality: a set of Romansh documents
|
|
| 54 |
|
| 55 |
Unit tests confirmed:
|
| 56 |
|
| 57 |
-
No false matches for known non-aligned documents
|
| 58 |
|
| 59 |
-
Low similarity scores for false-alignment candidates
|
| 60 |
|
| 61 |
## File Organization
|
| 62 |
|
|
@@ -64,4 +66,4 @@ Low similarity scores for false-alignment candidates
|
|
| 64 |
| --------------------- | ---------------------------------- |
|
| 65 |
| `parallel_data/` | All alignment outputs by threshold |
|
| 66 |
| `matched_0.600.jsonl` | Main result (highest quality) |
|
| 67 |
-
| `
|
|
|
|
| 15 |
|
| 16 |
threshold_matches/
|
| 17 |
|
| 18 |
+
├── matched_{THRESHOLD}_{ALPHA}.jsonl
|
| 19 |
|
| 20 |
+
├── matched_0.60_0.005.jsonl
|
| 21 |
|
| 22 |
...
|
| 23 |
|
| 24 |
+
With ALPHA being either 0.005 or None
|
| 25 |
+
|
| 26 |
|
| 27 |
## Format
|
| 28 |
|
|
|
|
| 56 |
|
| 57 |
Unit tests confirmed:
|
| 58 |
|
| 59 |
+
- No false matches (false positives) for known non-aligned documents
|
| 60 |
|
| 61 |
+
- Low similarity scores for false-alignment candidates
|
| 62 |
|
| 63 |
## File Organization
|
| 64 |
|
|
|
|
| 66 |
| --------------------- | ---------------------------------- |
|
| 67 |
| `parallel_data/` | All alignment outputs by threshold |
|
| 68 |
| `matched_0.600.jsonl` | Main result (highest quality) |
|
| 69 |
+
| `embedded` | German and Romansh embeddings |
|