Datasets:
File size: 4,463 Bytes
46e27b5 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 | """Corpus QA reporter — deterministic full re-verification of a committed dataset.
Single-label non-autoregressive speech disfluency contract:
1. Category must be CLEAN or DISFLUENCY.
2. All spans must have label "speech disfluency".
3. Zero targets in spans.
4. Clean speech emits spans: [].
5. Roundtrip slice-cut excision must exactly reconstruct clean_text.
6. Entity/Number conservation, pronoun safety, Tier-1 filler free.
"""
import argparse
import json
import re
import sys
from collections import Counter
sys.path.insert(0, "/opt/vox/sandbox/scripts")
from compiler import verify_roundtrip
from dedup import PROTECTED, TIER1, TIER1_ANY, near_dup, norm_tokens
from schema import DatasetExample
# Imported from dedup so the gate and the harvester can never drift apart.
SAFE_SPAN = PROTECTED
def main():
ap = argparse.ArgumentParser()
ap.add_argument("dataset")
ap.add_argument("--near-dup-jaccard", type=float, default=0.85)
ap.add_argument("--max-pairs", type=int, default=200000)
ap.add_argument("--report", default=None)
args = ap.parse_args()
fails = Counter()
examples = {}
ids, raws, fps = set(), set(), []
n = 0
for ln, line in enumerate(open(args.dataset), 1):
line = line.strip()
if not line:
continue
n += 1
try:
ex = DatasetExample(**{k: v for k, v in json.loads(line).items()
if k in DatasetExample.model_fields})
except Exception as e:
fails["1_schema"] += 1
examples.setdefault("1_schema", (ln, str(e)[:80]))
continue
if ex.category not in {"DISFLUENCY", "CLEAN"}:
fails["1_schema"] += 1
examples.setdefault("1_schema", (ln, f"Invalid category: {ex.category}"))
if ex.id in ids:
fails["2_id_unique"] += 1
examples.setdefault("2_id_unique", (ln, ex.id))
ids.add(ex.id)
key = ex.raw_text.strip().lower()
if key in raws:
fails["3_exact_unique"] += 1
raws.add(key)
fp = norm_tokens(ex.raw_text)
if near_dup(fp, fps, args.near_dup_jaccard):
fails["4_near_unique"] += 1
examples.setdefault("4_near_unique", (ln, ex.raw_text[:70]))
if len(fps) < args.max_pairs:
fps.append(fp)
if ex.raw_text.strip()[-1:] not in ".?!\u0964" or \
ex.clean_text.strip()[-1:] not in ".?!\u0964":
fails["10_boundary"] += 1
if ex.category == "CLEAN":
if ex.spans:
fails["5_clean_has_spans"] += 1
continue
for s in ex.spans:
if s.label != "speech disfluency":
fails["1_schema"] += 1
examples.setdefault("1_schema", (ln, f"Invalid span label: {s.label}"))
ok, rec = verify_roundtrip(ex.raw_text, ex.spans, ex.clean_text)
if not ok:
fails["5_roundtrip"] += 1
examples.setdefault("5_roundtrip", (ln, f"{rec!r} != {ex.clean_text!r}"))
continue
for s in ex.spans:
t = s.text.strip()
if TIER1.match(t) or TIER1_ANY.search(t):
fails["8_tier1_free"] += 1
examples.setdefault("8_tier1_free", (ln, t[:40]))
if t.lower().strip(",. ") in SAFE_SPAN:
fails["9_pronoun_safe"] += 1
examples.setdefault("9_pronoun_safe", (ln, t[:40]))
if TIER1_ANY.search(ex.clean_text):
fails["11_clean_not_tier1"] += 1
cats = Counter()
for line in open(args.dataset):
try:
cats[json.loads(line).get("category", "?")] += 1
except Exception:
pass
total = sum(fails.values())
print(f"dataset: {args.dataset}")
print(f"examples: {n} categories: {dict(cats)}")
if not fails:
print("QA REPORT: CLEAN — all deterministic checks pass")
else:
print(f"QA REPORT: {total} violation(s)")
for k in sorted(fails):
print(f" {k}: {fails[k]} e.g. {examples.get(k)}")
if args.report:
json.dump({"dataset": args.dataset, "examples": n,
"categories": dict(cats), "violations": dict(fails),
"examples_of_violation": {k: list(v) for k, v in examples.items()}},
open(args.report, "w"), indent=2)
sys.exit(1 if fails else 0)
if __name__ == "__main__":
main()
|