specfor commited on
Commit
12f8cb3
·
verified ·
1 Parent(s): 4c098a4

Leakage-free Sinhala/Pali/Sanskrit release

Browse files
Files changed (6) hide show
  1. .gitattributes +3 -0
  2. README.md +56 -0
  3. split_report.json +116 -0
  4. test.jsonl +3 -0
  5. train.jsonl +3 -0
  6. validation.jsonl +3 -0
.gitattributes CHANGED
@@ -58,3 +58,6 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
58
  # Video files - compressed
59
  *.mp4 filter=lfs diff=lfs merge=lfs -text
60
  *.webm filter=lfs diff=lfs merge=lfs -text
 
 
 
 
58
  # Video files - compressed
59
  *.mp4 filter=lfs diff=lfs merge=lfs -text
60
  *.webm filter=lfs diff=lfs merge=lfs -text
61
+ test.jsonl filter=lfs diff=lfs merge=lfs -text
62
+ train.jsonl filter=lfs diff=lfs merge=lfs -text
63
+ validation.jsonl filter=lfs diff=lfs merge=lfs -text
README.md ADDED
@@ -0,0 +1,56 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ language: [si, pi, sa]
3
+ license: other
4
+ task_categories: [text-classification]
5
+ pretty_name: Sinhala-script LangID (Sinhala, Pali, Sanskrit)
6
+ configs:
7
+ - config_name: default
8
+ data_files:
9
+ - {split: train, path: train.jsonl}
10
+ - {split: validation, path: validation.jsonl}
11
+ - {split: test, path: test.jsonl}
12
+ ---
13
+
14
+ # Sinhala-script language identification: Sinhala, Pali, Sanskrit
15
+
16
+ Sentence-level instances of three languages written in Sinhala script, with a
17
+ leakage-free document-blocked split. Labels: `sin_Sinh`, `pli_Sinh`, `san_Sinh`.
18
+
19
+ ## Rows
20
+
21
+ | split | sin_Sinh | pli_Sinh | san_Sinh | total |
22
+ |---|---:|---:|---:|---:|
23
+ | train | 83105 | 58309 | 10565 | 151979 |
24
+ | validation | 10345 | 7317 | 1337 | 18999 |
25
+ | test | 10567 | 7105 | 1310 | 18982 |
26
+
27
+ ## Construction
28
+
29
+ Built by `scripts/maintainer/resplit_target.py` (config: `target_split` in
30
+ `config/pipeline.yaml`):
31
+
32
+ 1. Sources pooled and NFC-normalised; text not in Sinhala script removed.
33
+ 2. Documents: source document ids; for the Pali-Sinhala parallel corpus
34
+ (sinhala-nlp/pali-sinhala, row-aligned) a sutta starts at each Pali incipit
35
+ *evam me sutam*, and both sides of a translation pair share the document.
36
+ 3. Split unit: contiguous block of <= 25 rows within a document.
37
+ 4. Sentence-level units: split at sentence terminators, packed to <= 200 characters.
38
+ 5. Exact duplicates removed; texts with conflicting labels removed.
39
+ 6. Near duplicates removed with MinHash LSH (character 5-grams, Jaccard >= 0.8).
40
+ 7. Group-stratified {'train': 0.8, 'validation': 0.1, 'test': 0.1} split (seed 42).
41
+
42
+ Leakage check: {'exact_cross_split': 0, 'near_dup_cross_split_j0.8': 0, 'near_dup_cross_split_j0.5': 15005}
43
+
44
+ Length (characters): {'pli_Sinh': {'p50': 138.0, 'p90': 195.0, 'max': 200}, 'san_Sinh': {'p50': 108.0, 'p90': 173.0, 'max': 200}, 'sin_Sinh': {'p50': 155.0, 'p90': 197.0, 'max': 200}}
45
+
46
+ ## Sources
47
+
48
+ - DCS: {'san_Sinh': 4888}
49
+ - SansinNT: {'san_Sinh': 8223}
50
+ - SiDiaC-v2: {'san_Sinh': 101, 'pli_Sinh': 2862}
51
+ - SiPaKosa: {'sin_Sinh': 5997}
52
+ - pali-sinhala-parallel: {'pli_Sinh': 69869, 'sin_Sinh': 98020}
53
+
54
+ ## Fields
55
+
56
+ `sample_id`, `text`, `label`, `source`, `subcorpus`, `doc_id`, `block_id`, `text_sha256`.
split_report.json ADDED
@@ -0,0 +1,116 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "inputs_sha256": {
3
+ "train.csv": "714e6df2d5cbcf54a74d9fb6851cc8c76cd70441bb6dc0ec6ba36a71a1f31a02",
4
+ "val.csv": "226bc6ccac463a462b8f4166d2110ab15a579c0130a337b392d969fddaa350ee",
5
+ "test.csv": "e8ff5efb050c395448d32af335507c5616d8bca0ef0c410d7f1f136ab0d9fd54"
6
+ },
7
+ "config": {
8
+ "seed": 42,
9
+ "fractions": {
10
+ "train": 0.8,
11
+ "validation": 0.1,
12
+ "test": 0.1
13
+ },
14
+ "max_chars": 200,
15
+ "min_letters": 5,
16
+ "block_size": 25,
17
+ "near_dup": {
18
+ "shingle": 5,
19
+ "num_perm": 128,
20
+ "threshold": 0.8
21
+ }
22
+ },
23
+ "stats": {
24
+ "input_rows": 74318,
25
+ "units_too_short_or_wrong_script": 219,
26
+ "rows_not_sinhala_script": 478,
27
+ "units_after_segmentation": 222146,
28
+ "exact_duplicate_removed": 13274,
29
+ "near_duplicate_removed": 18693,
30
+ "near_duplicate_pairs": 86443
31
+ },
32
+ "provenance": {
33
+ "public_corpus": "sinhala-nlp/pali-sinhala@da194116e510923936a3871a95ce284b5f75cde5",
34
+ "rows_checked": 52956,
35
+ "rows_matching_public_alignment": 52956
36
+ },
37
+ "parallel_documents": 441,
38
+ "splits": {
39
+ "train": {
40
+ "san_Sinh": {
41
+ "DCS": 3919,
42
+ "SansinNT": 6570,
43
+ "SiDiaC-v2": 76
44
+ },
45
+ "pli_Sinh": {
46
+ "SiDiaC-v2": 2297,
47
+ "pali-sinhala-parallel": 56012
48
+ },
49
+ "sin_Sinh": {
50
+ "SiPaKosa": 4804,
51
+ "pali-sinhala-parallel": 78301
52
+ }
53
+ },
54
+ "validation": {
55
+ "san_Sinh": {
56
+ "DCS": 483,
57
+ "SansinNT": 829,
58
+ "SiDiaC-v2": 25
59
+ },
60
+ "pli_Sinh": {
61
+ "SiDiaC-v2": 285,
62
+ "pali-sinhala-parallel": 7032
63
+ },
64
+ "sin_Sinh": {
65
+ "SiPaKosa": 604,
66
+ "pali-sinhala-parallel": 9741
67
+ }
68
+ },
69
+ "test": {
70
+ "san_Sinh": {
71
+ "DCS": 486,
72
+ "SansinNT": 824
73
+ },
74
+ "pli_Sinh": {
75
+ "SiDiaC-v2": 280,
76
+ "pali-sinhala-parallel": 6825
77
+ },
78
+ "sin_Sinh": {
79
+ "SiPaKosa": 589,
80
+ "pali-sinhala-parallel": 9978
81
+ }
82
+ }
83
+ },
84
+ "split_totals": {
85
+ "train": 151979,
86
+ "validation": 18999,
87
+ "test": 18982
88
+ },
89
+ "blocks_per_split": {
90
+ "train": 1985,
91
+ "validation": 236,
92
+ "test": 247
93
+ },
94
+ "length_chars": {
95
+ "pli_Sinh": {
96
+ "p50": 138.0,
97
+ "p90": 195.0,
98
+ "max": 200
99
+ },
100
+ "san_Sinh": {
101
+ "p50": 108.0,
102
+ "p90": 173.0,
103
+ "max": 200
104
+ },
105
+ "sin_Sinh": {
106
+ "p50": 155.0,
107
+ "p90": 197.0,
108
+ "max": 200
109
+ }
110
+ },
111
+ "leakage": {
112
+ "exact_cross_split": 0,
113
+ "near_dup_cross_split_j0.8": 0,
114
+ "near_dup_cross_split_j0.5": 15005
115
+ }
116
+ }
test.jsonl ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:073d6892de08d61df3ca04e30475863d48e772c070329f37785cea5bcc301926
3
+ size 13160201
train.jsonl ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ba705fe787d630d6ed2ed528042a9413d1eddabc48242be823cd03a0d24cf15e
3
+ size 105339396
validation.jsonl ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:08f3a52fb8d8b1f5036933dbdbbe84f0a4bd543900f2afc3b35b55f05b2a60a7
3
+ size 13179644