Release 1.0.2: conservative labelled country-area phone fix

CHANGELOG.md

@@ -8,6 +8,20 @@ Semver for this island:
 
 Accuracy is 841-dev, union at 0.95, 354 gold spans. A version that changes those numbers must update `hybrid.json` `eval` and the tables below.
 
+## 1.0.2
+
+Fix immediately labelled hyphenated country-area phones with one-digit country codes or wider area codes. Numeric continuations, slash lists, weak contact cues, and nearby title prose still abstain. Same weights, threshold, and API.
+
+841-dev: rules recover 14 more whole spans; the union recovers one (323 → 324), reducing residual passages 25 → 24. Rules FP remain 98; union FP remain 123. Lost gold, new false characters, and new clean-passage damage are all zero. Rules SHA `3016ae5b…`.
+
+Exact-span scores are recomputed from deduplicated raw rule/model span triples: precision 86.34%, recall 89.27%, F1 87.78%. The 1.0.1 card's precision/F1 were stale after the prefix trim changed exact rule/model duplicate counts; its character metrics were correct.
+
+| Version | Whole /354 | Residual | Rules FP | Union FP | Char P | Char R | What changed |
+|---|---:|---:|---:|---:|---:|---:|---|
+| 1.0.0 | 323 | 25 | 133 | 133 | 97.76% | 95.95% | First Hub snapshot |
+| 1.0.1 | 323 | 25 | 98 | 123 | 97.93% | 95.95% | Prefix-only glued-email trim |
+| 1.0.2 | 324 | 24 | 98 | 123 | 97.93% | 96.12% | Labelled country-area phone fix |
+
 ## 1.0.1
 
 Prefix-only glued-email trim. Cluster-gated title-case 5–11 letter prefixes are dropped from the redaction when the remainder is already a lowercase-local email. Same epoch-5 weights. Lost gold 0. New clean-passage damage 0.

README.md

@@ -13,7 +13,7 @@ tags:
 - hybrid
 ---
 
-# NERGAL 1.0.1
+# NERGAL 1.0.2
 
 **Named Entity Recognition with Grounded Additive Labels**
 
@@ -23,8 +23,8 @@ SlayerLab hybrid PII cleaner for Polish. Not a chat model. Not a drop-in `pipeli
 
 Python rules do the identifiers they can prove. A transformer NER head adds phone and other PII spans the regex misses. The cleaner **unions** the two on the original text, then replaces hits with `[Telefon]` or `[PII]`.
 
-- **Version:** `1.0.1` (`hybrid.json`, `CHANGELOG.md`)
-- **Ground:** `scrub_pii` regex (SHA256 `4dcc441c…`)
+- **Version:** `1.0.2` (`hybrid.json`, `CHANGELOG.md`)
+- **Ground:** `scrub_pii` regex (SHA256 `3016ae5b…`)
 - **Additive labels:** XLM-RoBERTa-large token classifier, BIO tags `phone` / `pii`, threshold 0.95
 - **This snapshot:** seed `202609160`, **epoch 5** of a seven-epoch schedule
 
@@ -35,7 +35,8 @@ Python rules do the identifiers they can prove. A transformer NER head adds phon
 | Version | Whole /354 | Residual | Rules FP | Union FP | Char P | Char R | What changed |
 |---|---:|---:|---:|---:|---:|---:|---|
 | 1.0.0 | 323 | 25 | 133 | 133 | 97.76% | 95.95% | First Hub snapshot |
-| **1.0.1** | **323** | **25** | **98** | **123** | **97.93%** | **95.95%** | Prefix-only glued-email trim |
+| 1.0.1 | 323 | 25 | 98 | 123 | 97.93% | 95.95% | Prefix-only glued-email trim |
+| **1.0.2** | **324** | **24** | **98** | **123** | **97.93%** | **96.12%** | Labelled country-area phone fix |
 
 ## 841-dev
 
@@ -71,23 +72,25 @@ Same 841-dev split, threshold **0.95**. **Naked** is the transformer alone. **
 
 | System | Mode | Whole /354 | Residual | False chars | Char P | Char R |
 |---|---|---:|---:|---:|---:|---:|
-| Regex (`scrub_pii`) | rules | 245 | 73 | 98 | 98.04% | 81.01% |
+| Regex (`scrub_pii`) | rules | 259 | 70 | 98 | 98.13% | 85.04% |
 | GLiNER 2.5-multi zero-shot | naked | 81 | 188 | 970 | 56.98% | 21.22% |
-| GLiNER 2.5-multi zero-shot | ∪ regex | 262 | 63 | 1,103 | 82.36% | 85.02% |
+| GLiNER 2.5-multi zero-shot | ∪ regex | 273 | 61 | 1,068 | 83.33% | 88.16% |
 | Historical GLiNER email12 | naked | 249 | 70 | 10 | 99.81% | 85.39% |
-| Historical GLiNER email12 | ∪ regex | 289 | 51 | 143 | 97.54% | 93.48% |
+| Historical GLiNER email12 | ∪ regex | 290 | 50 | 108 | 98.13% | 93.76% |
 | XLM-R epoch 5 | naked | 298 | 42 | 57 | 98.96% | 89.38% |
-| **NERGAL 1.0.1** | **∪ regex** | **323** | **25** | **123** | **97.93%** | **95.95%** |
+| **NERGAL 1.0.2** | **∪ regex** | 324 | 24 | 123 | 97.93% | 96.12% |
 
-Zero-shot [GLiNER 2.5](https://huggingface.co/fastino/gliner2.5-multi-v1) is not competitive here, especially on non-phone PII (7/185 whole vs 160 naked / 179 union). Fine-tuned historical GLiNER is the precise naked baseline (10 false characters) and still trails XLM-R on coverage. NERGAL is XLM-R epoch 5 plus the regex: 144/169 phone, 179/185 other PII. Exact-span precision 87.50%, recall 88.98%, F1 88.24%.
+Zero-shot [GLiNER 2.5](https://huggingface.co/fastino/gliner2.5-multi-v1) is not competitive here, especially on non-phone PII (7/185 whole vs 160 naked / 179 union). Fine-tuned historical GLiNER is the precise naked baseline (10 false characters) and still trails XLM-R on coverage. NERGAL is XLM-R epoch 5 plus the regex: 145/169 phone, 179/185 other PII. Exact-span precision 86.34%, recall 89.27%, F1 87.78%.
 
 Trained GLiNER, HerBERT-large, and XLM-R-large were also compared on this split (plot above). GLiNER’s 0.50 coverage lead is the precision collapse in that figure; no GLiNER or HerBERT operating point passed the content-preservation gate.
 
 ## Extra seeds
 
+Historical seed-comparison results, before the 1.0.2 parser fix.
+
 | Seed | Whole /354 | False chars | New false vs historical union |
 |---|---:|---:|---:|
-| 202609160 (this snapshot) | 323 | 123 | 0 |
+| 202609160 (selected weights) | 323 | 123 | 0 |
 | 202609161 | 322 | 134 | 1 |
 | 202609162 | 316 | 151 | 18 |
 
@@ -110,6 +113,6 @@ nergal = Nergal.from_pretrained(root, local_files_only=True)
 masked, counts = nergal.scrub(text)
 ```
 
-`hybrid.json` records version `1.0.1`, threshold 0.95, gap ids `250002` / `250003`, and the 841-dev `eval` block. `test_nergal.py` is synthetic (no corpus text). From this snapshot: `python -m unittest test_nergal`.
+`hybrid.json` records version `1.0.2`, threshold 0.95, gap ids `250002` / `250003`, and the 841-dev `eval` block. `test_nergal.py` is synthetic (no corpus text). From this snapshot: `python -m unittest test_nergal`.
 
 Base weights: [`FacebookAI/xlm-roberta-large`](https://huggingface.co/FacebookAI/xlm-roberta-large) revision `c23d21b0620b635a76227c604d44e43a9f0ee389` (MIT).

hybrid.json

@@ -1,25 +1,25 @@
 {
   "full_name": "Named Entity Recognition with Grounded Additive Labels",
   "hub_id": "SlayerLab/NERGAL",
-  "version": "1.0.1",
+  "version": "1.0.2",
   "mode": "rules_union",
   "epoch": 5,
   "seed": 202609160,
   "threshold": 0.95,
-  "rules_sha256": "4dcc441c39b653b1e8360928c9e36f6299c1f17e81e7566dd2d252cc5dbbb03c",
+  "rules_sha256": "3016ae5bd403ff997458f9dd74bad8c6ed1388eb83dadc1b31cdb182f9ed607f",
   "eval": {
     "split": "841-dev",
     "gold_entities": 354,
-    "whole_entities": 323,
-    "residual_passages": 25,
+    "whole_entities": 324,
+    "residual_passages": 24,
     "union_fp": 123,
     "rules_fp": 98,
     "character_precision": 0.9793,
-    "character_recall": 0.9595,
-    "exact_precision": 0.875,
-    "exact_recall": 0.8898,
-    "exact_f1": 0.8824,
-    "phone_whole": 144,
+    "character_recall": 0.9612,
+    "exact_precision": 0.8634,
+    "exact_recall": 0.8927,
+    "exact_f1": 0.8778,
+    "phone_whole": 145,
     "phone_gold": 169,
     "pii_whole": 179,
     "pii_gold": 185

nergal.py

@@ -17,13 +17,13 @@ import scrub_pii
 from scrub_pii import PHONE_TAG, PII_TAG
 
 HUB_ID = 'SlayerLab/NERGAL'
-VERSION = '1.0.1'
+VERSION = '1.0.2'
 GAPS = ['[PII_SPACE]', '[PII_BREAK]']
 GAP_IDS = [250002, 250003]
 BIO_LABELS = ['O', 'B-phone', 'I-phone', 'B-pii', 'I-pii']
 LABELS = ['phone', 'pii']
 THRESHOLD = 0.95
-RULES_SHA = '4dcc441c39b653b1e8360928c9e36f6299c1f17e81e7566dd2d252cc5dbbb03c'
+RULES_SHA = '3016ae5bd403ff997458f9dd74bad8c6ed1388eb83dadc1b31cdb182f9ed607f'
 
 
 def sha(path):

scrub_pii.py

@@ -21,7 +21,8 @@ unrelated lines end the list. Bounded staff/address-directory evidence also cove
 formatted phone fields. Labelled full-number ranges retain shared prefixes.
 Phones require a nearby contact cue, a Polish +48/0048 prefix, or explicit
 international country/trunk notation such as +CC (0). With a cue, the EUR-Lex
-"(32-2) 299 11 11" country-area form counts as international. Short service numbers
+"(32-2) 299 11 11" country-area form counts as international. Strong labels also
+admit one-digit country codes and wider hyphenated area codes. Short service numbers
 need strong labels; 116xxx numbers also accept nearby telephone prose. Unlabelled
 domestic numbers are left for audit because table cells have the same shapes.
 Flattened tables can glue labels to both neighbours ("Mödlingtel.: … 38112faks:");
@@ -139,7 +140,7 @@ _GLUED_PHONE_LABEL = r"(?:tel\.|telefon|t[ée]l[ée]phone|t[ée]l[ée]copieur|(?
 _PHONE_END = r"(?:(?!\w)|(?=" + _GLUED_PHONE_LABEL + r"|(?:e-?mail|t[ée]lex)[ \t]*:))"
 _PHONE_RE = re.compile(
     _VANITY_PHONE_RE.pattern + r"|(?:(?<!\w)|(?<=telefon)|(?<=tel)|(?<=faks)|(?<=fax))(?:\(?\+[ \t]*)?"
-    r"(?:\((?:0|0?\d{2,4}(?:-\d{1,2})?)\)" + _SEP + r")?"
+    r"(?:\((?:0|[1-9]\d{0,2}-\d{1,4}|0?\d{2,4}(?:-\d{1,2})?)\)" + _SEP + r")?"
     r"\d(?:" + _SEP + r"\d){4,}" + _PHONE_END, re.I,
 )
 _SHORT_PHONE_LABEL_RE = re.compile(
@@ -179,6 +180,7 @@ _PHONE_HEADING_RE = re.compile(
     r"(?:tj\.|nr\.|[^\d\n.!?;\[\]]){0,75}:?[ \t]*(?:\n[ \t]*){1,4}\Z", re.I)
 _CONTACT_EXCLUDE_RE = re.compile(r"\b(?:spraw\w*|kod\w*|kwot\w*|statystyk\w*|taryf\w*)\b", re.I)
 _PHONE_LINE_RE = re.compile(r"(?<!\w)\d{2,3}(?:[ \t-]\d{2,4}){2,3}(?!\w)")
+_PHONE_TITLE_RE = re.compile(r"\b(?:film\w*|serial\w*|tytuł\w*|zatytułowan\w*)\b", re.I)
 _STAFF_ROLE_RE = re.compile(
     r"\b(?:inspektor|referent|koordynator|psycholog|księgowość|księgowy|księgowa|"
     r"pracownicy[ \t]+socjalni|łowczy|podłowczy)\b", re.I)
@@ -484,7 +486,7 @@ def _phone_ok(raw: str, short: bool = False) -> bool:
     if short and re.match(r'^\(0?\d{2,4}(?:-\d{1,2})?\)', s):
         return 8 <= len(d) <= 15
     # EUR-Lex puts country and area code in parentheses: "(32-2) 299 11 11" is +32 2 299 11 11.
-    s = re.sub(r'^\(([1-9]\d{1,2})-(\d{1,2})\)', r'+\1 \2', s)
+    s = re.sub(r'^\(([1-9]\d{0,2})-(\d{1,4})\)', r'+\1 \2', s)
     if s.startswith(('+', '00')):
         international = _digits(s.replace('(0)', ''))
         if s.startswith('00'):
@@ -595,10 +597,21 @@ def _replace_phones(text: str, *, mask=_tag) -> tuple[str, int]:
                 out.append(text[pos:line_end+1])
                 pos = line_end+1
                 continue
-        short = bool(_SHORT_PHONE_LABEL_RE.search(text[max(0, m.start()-100):m.start()]))
+        immediate_phone_label = bool(_SHORT_PHONE_LABEL_RE.search(text[max(0, m.start()-100):m.start()]))
+        short = immediate_phone_label
         short = short or (re.fullmatch(r'116[ \t]?\d{3}', raw)
                           and _phone_context(text, m.start(), raw))
         short = short or _phone_list_context(text, m.start())
+        novel_country_area = (bool(re.match(r'^\([1-9]\d{0,2}-\d{1,4}\)', raw))
+                              and not re.match(r'^\(0?\d{2,4}(?:-\d{1,2})?\)', raw))
+        # Broadened country-area forms require an immediate strong label and
+        # reject numeric continuations, slash lists, and nearby title prose.
+        novel_country_area_ok = not novel_country_area or (
+            immediate_phone_label
+            and not re.search(r'\r?\n[ \t]*\d|/', raw)
+            and not re.match(r'[ \t]*(?:\r?\n[ \t]*\d|/)', text[end:])
+            and not _PHONE_TITLE_RE.search(text[max(0, m.start()-100):m.start()])
+        )
         # A bare five-digit continuation followed by a place/name can be a
         # postal address. It needs its own phone label to override that ambiguity.
         postal = re.fullmatch(r'\d{5}', raw) and re.match(r'[ \t]+[^\W\d_]', text[end:])
@@ -617,8 +630,8 @@ def _replace_phones(text: str, *, mask=_tag) -> tuple[str, int]:
                                  text[last_phone_end:m.start()], re.I)
                 and _pl_national_ok(_digits(raw))):
             short = True
-        if (short or _phone_context(text, m.start(), raw)
-                or _directory_context(text, m.start(), line_end if line_end >= 0 else end)):
+        if (novel_country_area_ok and (short or _phone_context(text, m.start(), raw)
+                or _directory_context(text, m.start(), line_end if line_end >= 0 else end))):
             if (_is_amount(text, m.start(), end)
                     and re.fullmatch(r'\d{1,3}(?:\.[ \t]*\d{3})+', raw)):
                 out.extend((text[pos:m.start()], raw))

test_nergal.py

@@ -5,7 +5,7 @@ import unittest
 from pathlib import Path
 
 HERE = Path(__file__).resolve().parent
-RULES_SHA = '4dcc441c39b653b1e8360928c9e36f6299c1f17e81e7566dd2d252cc5dbbb03c'
+RULES_SHA = '3016ae5bd403ff997458f9dd74bad8c6ed1388eb83dadc1b31cdb182f9ed607f'
 
 
 class NergalTests(unittest.TestCase):
@@ -13,7 +13,7 @@ class NergalTests(unittest.TestCase):
         from nergal import GAP_IDS, GAPS, HUB_ID, RULES_SHA as PINNED, THRESHOLD, VERSION
         card = json.loads((HERE / 'hybrid.json').read_text())
         self.assertEqual(HUB_ID, 'SlayerLab/NERGAL')
-        self.assertEqual(VERSION, '1.0.1')
+        self.assertEqual(VERSION, '1.0.2')
         self.assertEqual(card['version'], VERSION)
         self.assertEqual(card['eval']['union_fp'], 123)
         self.assertEqual(card['eval']['rules_fp'], 98)