@stll/anonymize-wasm 1.4.1 → 1.4.4

package/dist/wasm.mjs

@@ -2,7 +2,7 @@ import { DEFAULT_ENTITY_LABELS, DETECTION_SOURCES, DETECTOR_PRIORITY, OPERATOR_T
 import { t as amount_words_default } from "./amount-words.mjs";
 import { t as address_street_types_default } from "./address-street-types.mjs";
 import { TextSearch } from "@stll/text-search-wasm";
-import { at, be, bg, br, cy, cz, de, dk, ee, es, fi, fr, gb, gr, hr, hu, ie, it, lt, lu, lv, mt, nl, pl, pt, ro, se, si, sk } from "@stll/stdnum";
+import { at, au, be, bg, br, ch, cy, cz, de, dk, ee, es, fi, fr, gb, gr, hr, hu, ie, it, lt, lu, lv, mt, nl, no, pl, pt, ro, se, si, sk, us } from "@stll/stdnum";
 import { toRegex } from "@stll/stdnum/patterns";
 //#region src/search-engine.ts
 let _TextSearch;
@@ -31,6 +31,7 @@ const createPipelineContext = () => ({
 	searchKey: "",
 	searchPromise: null,
 	nameCorpus: null,
+	nameCorpusKey: "",
 	nameCorpusPromise: null,
 	stopwords: null,
 	stopwordsPromise: null,
@@ -230,12 +231,22 @@ const LEGAL_SUFFIXES = [...[
 	"OHG",
 	"Ltd.",
 	"Ltd",
+	"LTD.",
+	"LTD",
 	"LLC",
 	"LLP",
 	"Inc.",
+	"INC.",
+	"Inc",
+	"INC",
 	"Corp.",
+	"CORP.",
+	"Corp",
+	"CORP",
 	"Corporation",
+	"CORPORATION",
 	"Co.",
+	"CO.",
 	"LP",
 	"L.P.",
 	"PLC",
@@ -245,6 +256,8 @@ const LEGAL_SUFFIXES = [...[
 	"B.V.",
 	"Pty Ltd.",
 	"Pty Ltd",
+	"PTY LTD.",
+	"PTY LTD",
 	"S.A.",
 	"SA",
 	"SAS",
@@ -889,7 +902,44 @@ const LOWER = "a-záčďéěíňóřšťúůýžäöüßàâæçèêëîïôùû
 const CAP_WORD = `(?:[${UPPER}]{2,}|[${UPPER}][${LOWER}${UPPER}]+)`;
 const SINGLE_CAP = `[${UPPER}](?![${LOWER}${UPPER}])`;
 const ALLCAP_WORD = `[${UPPER}]{2,}`;
+const HSPACE = "(?:[^\\S\\n]|[\xA0 ])";
+const LEGAL_LIST_BOUNDARY_RE = new RegExp(`^[,;]${HSPACE}+(?=\\p{Lu}|(?:\\p{Lu}\\.${HSPACE}?){2,})`, "u");
 const ROMAN_NUMERAL_RE = /^(?=[IVXLCDM])M{0,3}(?:CM|CD|D?C{0,3})(?:XC|XL|L?X{0,3})(?:IX|IV|V?I{0,3})$/;
+const EMPTY_LEADING_CLAUSE_TRIMS = {
+	phrases: [],
+	directPrefixes: []
+};
+let leadingClauseTrimsCache = null;
+let leadingClauseTrimsPromise = null;
+const loadLeadingClauseTrims = async () => {
+	if (leadingClauseTrimsCache) return leadingClauseTrimsCache;
+	if (leadingClauseTrimsPromise) return leadingClauseTrimsPromise;
+	leadingClauseTrimsPromise = (async () => {
+		let data = {};
+		try {
+			const mod = await import("./legal-form-leading-clauses.mjs");
+			data = mod.default ?? mod;
+		} catch (err) {
+			console.warn("[anonymize] legal-forms: failed to load legal-form-leading-clauses.json:", err);
+		}
+		const phrases = /* @__PURE__ */ new Set();
+		const directPrefixes = /* @__PURE__ */ new Set();
+		for (const [key, value] of Object.entries(data)) {
+			if (key.startsWith("_") || typeof value !== "object" || value === null) continue;
+			const config = value;
+			for (const phrase of config.phrases ?? []) if (typeof phrase === "string" && phrase.length > 0) phrases.add(phrase);
+			for (const prefix of config.directPrefixes ?? []) if (typeof prefix === "string" && prefix.length > 0) directPrefixes.add(prefix);
+		}
+		const result = {
+			phrases: [...phrases],
+			directPrefixes: [...directPrefixes]
+		};
+		leadingClauseTrimsCache = result;
+		return result;
+	})();
+	return leadingClauseTrimsPromise;
+};
+const getLeadingClauseTrimsSync = () => leadingClauseTrimsCache ?? EMPTY_LEADING_CLAUSE_TRIMS;
 let legalRoleHeadsCache = null;
 let legalRoleHeadsPromise = null;
 const loadLegalRoleHeads = async () => {
@@ -917,7 +967,8 @@ const warmLegalRoleHeads = async () => {
 		loadSentenceVerbIndicators(),
 		loadClauseNounHeads(),
 		loadConnectorProseHeads(),
-		loadStructuralSingleCapPrefixes()
+		loadStructuralSingleCapPrefixes(),
+		loadLeadingClauseTrims()
 	]);
 };
 let allLegalSuffixesCache = null;
@@ -1054,13 +1105,12 @@ const loadStructuralSingleCapPrefixes = async () => {
 	return structuralSingleCapPrefixesPromise;
 };
 const getStructuralSingleCapPrefixesSync = () => structuralSingleCapPrefixesCache ?? /* @__PURE__ */ new Set();
-const escapeForRegex = (form) => form.replace(/[.*+?^${}()|[\]\\]/g, "\\$&").replace(/\s+/g, "\\s+").replace(/\\\./g, "\\.[^\\S\\n]?");
+const escapeForRegex = (form) => form.replace(/[.*+?^${}()|[\]\\]/g, "\\$&").replace(/\s+/g, `${HSPACE}+`).replace(/\\\./g, `\\.${HSPACE}?`);
 const isShortForm = (form) => form.replace(/[.\s]/g, "").length <= 3 && !form.includes(" ");
 const buildDottedAbbreviationAlternation = (forms) => [...new Set(forms.filter((form) => /^[\p{Lu}][\p{L}\p{M}]{0,5}\.$/u.test(form)).map((form) => form.slice(0, -1)).filter((form) => form.length > 0))].toSorted((a, b) => b.length - a.length).map(escapeForRegex).join("|");
 const buildPatternString = (forms) => {
 	if (forms.length === 0) return null;
 	const alt = forms.toSorted((a, b) => b.length - a.length).map(escapeForRegex).join("|");
-	const HSPACE = "[^\\S\\n]";
 	const LOWER_CONNECTOR = `${HSPACE}+(?:a|and|und|et|e|y|i)${HSPACE}+(?=[${LOWER}])`;
 	const SIMPLE_SEP = `(?:${HSPACE}|[&,.${DASH_INNER}]){1,4}`;
 	const CAP_OR_NUM_WORD = `(?:${CAP_WORD}|${SINGLE_CAP}|\\d{1,4})`;
@@ -1068,7 +1118,7 @@ const buildPatternString = (forms) => {
 	const ANY_WORD_TAIL = `(?:(?!(?:and|und|et)(?![${UPPER}${LOWER}]))[${UPPER}${LOWER}][${LOWER}${UPPER}]+|[${UPPER}]{2,3}|\\d{1,4})`;
 	const dottedAbbreviationAlt = buildDottedAbbreviationAlternation(forms);
 	const dottedAbbreviationTail = dottedAbbreviationAlt.length > 0 ? `(?:${SIMPLE_SEP}(?:${dottedAbbreviationAlt})\\.)?` : "";
-	return `${`(?:${`(?:${CAP_WORD})(?:${SIMPLE_SEP}(?:${CAP_OR_NUM_WORD})){0,10}` + dottedAbbreviationTail})(?:${`${SIMPLE_SEP}(?:${LOWER_WORD})(?:(?:${LOWER_CONNECTOR}|${SIMPLE_SEP})(?:${ANY_WORD_TAIL})){0,10}`})?`}(?:\\s+|,\\s*)(?:${alt})(?![${LOWER}])`;
+	return `${`(?:${`(?:${CAP_WORD})(?:${SIMPLE_SEP}(?:${CAP_OR_NUM_WORD})){0,10}` + dottedAbbreviationTail})(?:${`${SIMPLE_SEP}(?:${LOWER_WORD})(?:(?:${LOWER_CONNECTOR}|${SIMPLE_SEP})(?:${ANY_WORD_TAIL})){0,10}`})?`}${`(?:${HSPACE}+|,${HSPACE}*)`}(?:${alt})(?![${LOWER}])`;
 };
 /**
 * Build legal form regex pattern strings.
@@ -1098,10 +1148,17 @@ const buildLegalFormPatterns = async () => {
 	if (longPattern) patterns.push(longPattern);
 	const shortPattern = buildPatternString(allForms.filter(isShortForm));
 	if (shortPattern) patterns.push(shortPattern);
-	const allcapPrefix = `(?:${ALLCAP_WORD})(?:[ \\t&,.${DASH_INNER}]{1,4}(?:${ALLCAP_WORD})){0,2}`;
 	const allcapAlt = allForms.toSorted((a, b) => b.length - a.length).map(escapeForRegex).join("|");
-	patterns.push(`${allcapPrefix}(?:[ \\t]+|,[ \\t]*)(?:${allcapAlt})(?![${LOWER}])`);
-	patterns.push(`(?:^|(?<=[^${UPPER}${LOWER}\\p{N}]))[${UPPER}](?:[ \\t]+|,[ \\t]*)(?:${allcapAlt})(?![${UPPER}${LOWER}\\p{N}])`);
+	const mixedNameSep = `(?:${HSPACE}|[&,.${DASH_INNER}]){1,4}`;
+	const capitalizedNoDigitPrefix = `(?:${CAP_WORD})(?:${mixedNameSep}(?:${CAP_WORD}|${ALLCAP_WORD})){1,8}`;
+	patterns.push(`${capitalizedNoDigitPrefix}(?:${HSPACE}+|,${HSPACE}*)(?:${allcapAlt})(?![${LOWER}])`);
+	const allcapMixedPrefix = `(?:${ALLCAP_WORD})(?:${mixedNameSep}(?:${CAP_WORD}|${ALLCAP_WORD}|\\d{1,4})){1,6}`;
+	patterns.push(`${allcapMixedPrefix}(?:${HSPACE}+|,${HSPACE}*)(?:${allcapAlt})(?![${LOWER}])`);
+	const dottedLineWrapPrefix = `(?:${CAP_WORD})(?:${mixedNameSep}(?:${CAP_WORD}|${ALLCAP_WORD})){1,8}\\.${HSPACE}*\\n${HSPACE}*`;
+	patterns.push(`${dottedLineWrapPrefix}(?:${allcapAlt})(?![${LOWER}])`);
+	const allcapPrefix = `(?:${ALLCAP_WORD})(?:(?:${HSPACE}|[&,.${DASH_INNER}]){1,4}(?:${ALLCAP_WORD})){0,2}`;
+	patterns.push(`${allcapPrefix}(?:${HSPACE}+|,${HSPACE}*)(?:${allcapAlt})(?![${LOWER}])`);
+	patterns.push(`(?:^|(?<=[^${UPPER}${LOWER}\\p{N}]))[${UPPER}](?:${HSPACE}+|,${HSPACE}*)(?:${allcapAlt})(?![${UPPER}${LOWER}\\p{N}])`);
 	return patterns;
 };
 const CONNECTOR_RE = /^(?:a|and|und|et|e|y|i|&)$/i;
@@ -1110,13 +1167,14 @@ const UPPER_LETTER_RE = /^\p{Lu}/u;
 const COMPANY_SUFFIX_WORDS_RE = /^(?:Company|Co|Bank|Brothers|Bros|Sons|Group|Holdings|Trust|Partners|Associates|Corporation|Industries|Enterprises|Solutions|Systems|Services|Foundation|Institute)$/;
 const IN_NAME_PREPOSITION_RE = /^(?:of|the)$/i;
 const ENTITY_HEAD_WORD_RE = /^[\p{L}\p{M}&]+/u;
-const LEADING_CLAUSE_RE = /(?:^|\s)(?:by\s+and\s+between|is\s+between)\s+/giu;
-const BARE_SINGLE_CAP_LEGAL_FORM_RE = new RegExp(`^[${UPPER}](?:[ \\t]+|,[ \\t]*)`, "u");
-const STRUCTURAL_SINGLE_CAP_RE = new RegExp(`^([\\p{L}\\p{M}]+)[ \\t]+[${UPPER}](?:[.${DASH_INNER}]?\\d{1,3})?(?:[ \\t]+|,[ \\t]*)`, "u");
+const BARE_SINGLE_CAP_LEGAL_FORM_RE = new RegExp(`^[${UPPER}](?:${HSPACE}+|,${HSPACE}*)`, "u");
+const STRUCTURAL_SINGLE_CAP_RE = new RegExp(`^([\\p{L}\\p{M}]+)${HSPACE}+[${UPPER}](?:[.${DASH_INNER}]?\\d{1,3})?(?:${HSPACE}+|,${HSPACE}*)`, "u");
 const isStructuralSingleCapMatch = (text) => {
 	const first = STRUCTURAL_SINGLE_CAP_RE.exec(text)?.[1];
 	return first !== void 0 && getStructuralSingleCapPrefixesSync().has(first.toLowerCase());
 };
+const findLastSuffixSeparator = (text) => Math.max(text.lastIndexOf(" "), text.lastIndexOf("	"), text.lastIndexOf("\xA0"), text.lastIndexOf(" "), text.lastIndexOf(","));
+const stripDocxSpaces = (text) => text.replace(/[  ]/g, "");
 /**
 * Find the word ending just before `pos` in `text`,
 * skipping any whitespace (not newlines).
@@ -1180,6 +1238,58 @@ const trimEmbeddedLegalFormListPrefix = (entityStart, entityText) => {
 		entityText: entityText.slice(cut)
 	};
 };
+const splitEmbeddedLegalFormList = (entityStart, entityText) => {
+	const cuts = [0];
+	for (const suffix of getAllLegalSuffixesSync()) {
+		const suffixClean = suffix.replace(/[.,\s]/g, "");
+		if (suffixClean.length > 0 && ROMAN_NUMERAL_RE.test(suffixClean)) continue;
+		let fromIndex = 0;
+		while (fromIndex < entityText.length) {
+			const suffixStart = entityText.indexOf(suffix, fromIndex);
+			if (suffixStart === -1) break;
+			fromIndex = suffixStart + suffix.length;
+			const suffixEnd = suffixStart + suffix.length;
+			if (suffixEnd >= entityText.length - 1) continue;
+			const afterSuffix = entityText.slice(suffixEnd);
+			const boundary = LEGAL_LIST_BOUNDARY_RE.exec(afterSuffix);
+			if (boundary === null) continue;
+			const nextStart = suffixEnd + boundary[0].length;
+			cuts.push(nextStart);
+		}
+	}
+	const uniqueCuts = [...new Set(cuts)].toSorted((a, b) => a - b);
+	if (uniqueCuts.length === 1) return [{
+		entityStart,
+		entityText
+	}];
+	const segments = [];
+	for (let index = 0; index < uniqueCuts.length; index++) {
+		const start = uniqueCuts[index];
+		const end = uniqueCuts[index + 1] ?? entityText.length;
+		if (start === void 0) continue;
+		const segmentText = entityText.slice(start, end).replace(/[,\s;]+$/u, "");
+		if (segmentText.length === 0) continue;
+		if (!getAllLegalSuffixesSync().some((form) => segmentText.endsWith(form))) continue;
+		segments.push({
+			entityStart: entityStart + start,
+			entityText: segmentText
+		});
+	}
+	return segments;
+};
+const hasDisallowedLineBreak = (text) => {
+	for (const match of text.matchAll(/\n/gu)) {
+		const index = match.index;
+		if (index === void 0) continue;
+		const before = text.slice(0, index);
+		const after = text.slice(index + 1);
+		const dottedDesignatorBefore = /\.[^\S\n]*$/u.test(before);
+		const legalSuffixAfter = /^[^\S\n]*(?:\p{Lu}\.[^\S\n]?){1,}\p{Lu}?\.?$/u.test(after);
+		const allCapsSuffixAfter = /^[^\S\n]*\p{Lu}{2,}\.?$/u.test(after);
+		if (!dottedDesignatorBefore || !legalSuffixAfter && !allCapsSuffixAfter) return true;
+	}
+	return false;
+};
 const hasMiddleInitialBefore = (fullText, pos) => {
 	const previousWord = findWordBefore(fullText, pos);
 	if (!previousWord) return false;
@@ -1336,7 +1446,37 @@ const extendBackward = (fullText, matchStart, options = {}) => {
 };
 const trimLeadingClause = (text) => {
 	let cut = -1;
-	for (const match of text.matchAll(LEADING_CLAUSE_RE)) cut = match.index + match[0].length;
+	const trims = getLeadingClauseTrimsSync();
+	const phraseAlternation = trims.phrases.map(escapeForRegex).join("|");
+	if (phraseAlternation.length > 0) {
+		const phraseRe = new RegExp(`(?:^|\\s)(?:${phraseAlternation})${HSPACE}+`, "giu");
+		for (const match of text.matchAll(phraseRe)) cut = Math.max(cut, match.index + match[0].length);
+	}
+	const directPrefixAlternation = trims.directPrefixes.map(escapeForRegex).join("|");
+	const COMMA_GATED_DIRECT_PREFIXES = new Set([
+		"among",
+		"amongst",
+		"between"
+	]);
+	if (directPrefixAlternation.length > 0) {
+		const directPrefixRe = new RegExp(`\\b(?:${directPrefixAlternation})${HSPACE}+(?=\\p{Lu})`, "giu");
+		for (const match of text.matchAll(directPrefixRe)) {
+			const matchedPrefix = match[0].trim().toLowerCase();
+			const before = text.slice(0, match.index);
+			if (COMMA_GATED_DIRECT_PREFIXES.has(matchedPrefix) && !/,\s*$/u.test(before)) continue;
+			const words = before.match(/\p{L}[\p{L}\p{M}]*/gu) ?? [];
+			if (words.length >= 3 && words.some((word) => /\p{Ll}/u.test(word))) cut = Math.max(cut, match.index + match[0].length);
+		}
+	}
+	for (const match of text.matchAll(/,/gu)) {
+		const comma = match.index;
+		if (comma === void 0) continue;
+		const before = text.slice(0, comma);
+		if (!/\d/u.test(before)) continue;
+		const after = text.slice(comma + 1);
+		const leadingWs = after.match(/^\s*/u)?.[0].length ?? 0;
+		if ((after.slice(leadingWs).match(/\p{Lu}[\p{L}\p{M}\p{N}]*/gu) ?? []).length >= 3) cut = Math.max(cut, comma + 1 + leadingWs);
+	}
 	if (cut <= 0) return {
 		offset: 0,
 		text
@@ -1419,7 +1559,7 @@ const processLegalFormMatches = (allMatches, sliceStart, sliceEnd, fullText) =>
 				}
 			}
 		}
-		if (processedText.includes("\n")) continue;
+		if (processedText.includes("\n") && hasDisallowedLineBreak(processedText)) continue;
 		let entityStart = processedStart;
 		let entityText = processedText;
 		if (fullText && !trimmed) {
@@ -1429,49 +1569,49 @@ const processLegalFormMatches = (allMatches, sliceStart, sliceEnd, fullText) =>
 				entityText = fullText.slice(extended, processedStart + processedText.length).trimEnd();
 			}
 		}
-		const listTrim = trimEmbeddedLegalFormListPrefix(entityStart, entityText);
-		entityStart = listTrim.entityStart;
-		entityText = listTrim.entityText;
-		const clauseTrim = trimLeadingClause(entityText);
-		if (clauseTrim.offset > 0) {
-			entityStart += clauseTrim.offset;
-			entityText = clauseTrim.text;
-		}
-		const getPrefixInfo = (value) => {
-			const prefixEnd = value.lastIndexOf(",") !== -1 ? value.lastIndexOf(",") : value.lastIndexOf(" ");
-			return {
-				prefixEnd,
-				prefixPart: prefixEnd > 0 ? value.slice(0, prefixEnd).replace(/[^a-zA-ZÀ-ž]/g, "") : value.replace(/[^a-zA-ZÀ-ž]/g, "")
-			};
-		};
-		let { prefixEnd, prefixPart } = getPrefixInfo(entityText);
-		let isAllCapsMatch = prefixPart.length > 2 && prefixPart === prefixPart.toUpperCase();
-		if (isAllCapsMatch && fullText) {
-			const lineStart = fullText.lastIndexOf("\n", entityStart);
-			const lineEnd = fullText.indexOf("\n", entityStart + entityText.length);
-			const lineLetters = fullText.slice(lineStart + 1, lineEnd === -1 ? fullText.length : lineEnd).replace(/[^a-zA-ZÀ-ž]/g, "");
-			const upperCount = [...lineLetters].filter((c) => c === c.toUpperCase()).length;
-			if (lineLetters.length > 5 && upperCount / lineLetters.length >= .95) continue;
-			if ((prefixPart.length > 0 ? entityText.slice(0, prefixEnd > 0 ? prefixEnd : entityText.length).trim().split(/\s+/).length : 0) > 3) {
-				entityStart = match.start;
-				entityText = text;
-				({prefixEnd, prefixPart} = getPrefixInfo(entityText));
-				isAllCapsMatch = prefixPart.length > 2 && prefixPart === prefixPart.toUpperCase();
+		for (const segment of splitEmbeddedLegalFormList(entityStart, entityText)) {
+			entityStart = segment.entityStart;
+			entityText = segment.entityText;
+			const listTrim = trimEmbeddedLegalFormListPrefix(entityStart, entityText);
+			entityStart = listTrim.entityStart;
+			entityText = listTrim.entityText;
+			const clauseTrim = trimLeadingClause(entityText);
+			if (clauseTrim.offset > 0) {
+				entityStart += clauseTrim.offset;
+				entityText = clauseTrim.text;
 			}
-		} else if (isAllCapsMatch) continue;
-		const lastSpace = entityText.lastIndexOf(" ");
-		const rawSuffix = lastSpace !== -1 ? entityText.slice(lastSpace + 1) : "";
-		const suffixClean = rawSuffix.replace(/[.,]/g, "");
-		if (suffixClean.length > 0 && ROMAN_NUMERAL_RE.test(suffixClean)) continue;
-		if (suffixClean.length <= 2 && !/\./.test(rawSuffix) && /[^\x00-\x7F]/.test(entityText.slice(0, lastSpace !== -1 ? lastSpace : entityText.length))) continue;
-		results.push({
-			start: entityStart,
-			end: entityStart + entityText.length,
-			label: "organization",
-			text: entityText,
-			score: .95,
-			source: DETECTION_SOURCES.LEGAL_FORM
-		});
+			if (entityText.includes("\n") && hasDisallowedLineBreak(entityText)) continue;
+			const getPrefixInfo = (value) => {
+				const prefixEnd = value.lastIndexOf(",") !== -1 ? value.lastIndexOf(",") : value.lastIndexOf(" ");
+				return {
+					prefixEnd,
+					prefixPart: prefixEnd > 0 ? value.slice(0, prefixEnd).replace(/[^a-zA-ZÀ-ž]/g, "") : value.replace(/[^a-zA-ZÀ-ž]/g, "")
+				};
+			};
+			let { prefixEnd, prefixPart } = getPrefixInfo(entityText);
+			let isAllCapsMatch = prefixPart.length > 2 && prefixPart === prefixPart.toUpperCase();
+			if (isAllCapsMatch && fullText) {
+				if ((prefixPart.length > 0 ? entityText.slice(0, prefixEnd > 0 ? prefixEnd : entityText.length).trim().split(/\s+/).length : 0) > 3) {
+					entityStart = match.start;
+					entityText = text;
+					({prefixEnd, prefixPart} = getPrefixInfo(entityText));
+					isAllCapsMatch = prefixPart.length > 2 && prefixPart === prefixPart.toUpperCase();
+				}
+			} else if (isAllCapsMatch) continue;
+			const lastSuffixSeparator = findLastSuffixSeparator(entityText);
+			const rawSuffix = lastSuffixSeparator !== -1 ? entityText.slice(lastSuffixSeparator + 1) : "";
+			const suffixClean = rawSuffix.replace(/[.,]/g, "");
+			if (suffixClean.length > 0 && ROMAN_NUMERAL_RE.test(suffixClean)) continue;
+			if (suffixClean.length <= 2 && !/\./.test(rawSuffix) && /[^\x00-\x7F]/.test(stripDocxSpaces(entityText.slice(0, lastSuffixSeparator !== -1 ? lastSuffixSeparator : entityText.length)))) continue;
+			results.push({
+				start: entityStart,
+				end: entityStart + entityText.length,
+				label: "organization",
+				text: entityText,
+				score: .95,
+				source: DETECTION_SOURCES.LEGAL_FORM
+			});
+		}
 	}
 	return results;
 };
@@ -1885,20 +2025,31 @@ const getNameCorpusTitles = (ctx = defaultContext) => ctx.nameCorpus?.titlesList
 *   with per-language first names and surnames. When
 *   omitted, only legacy config files are used.
 */
-const initNameCorpus = (ctx = defaultContext, dictionaries) => {
-	if (ctx.nameCorpusPromise) return ctx.nameCorpusPromise;
+const initNameCorpus = (ctx = defaultContext, dictionaries, languages) => {
+	const languageKey = languages?.toSorted().join(",") ?? "*";
+	if (ctx.nameCorpus && ctx.nameCorpusKey === languageKey) return Promise.resolve();
+	if (ctx.nameCorpusPromise && ctx.nameCorpusKey === languageKey) return ctx.nameCorpusPromise;
+	ctx.nameCorpus = null;
+	ctx.nameCorpusKey = languageKey;
 	const promise = (async () => {
 		try {
-			const [legacyFirstMod, legacySurnameMod, titleMod, exclusionMod] = await Promise.all([
+			const [legacyFirstMod, legacySurnameMod, titleMod, exclusionMod, commonWordsMod] = await Promise.all([
 				import("./names-first.mjs"),
 				import("./names-surnames.mjs"),
 				import("./names-title-tokens.mjs"),
-				import("./names-exclusions.mjs")
+				import("./names-exclusions.mjs"),
+				import("./common-words-en.mjs")
 			]);
 			const firstNames = [...legacyFirstMod.default.names];
-			if (dictionaries?.firstNames) for (const names of Object.values(dictionaries.firstNames)) for (const name of names) firstNames.push(name);
+			if (dictionaries?.firstNames) {
+				const entries = languages === void 0 ? Object.entries(dictionaries.firstNames) : Object.entries(dictionaries.firstNames).filter(([language]) => languages.includes(language));
+				for (const [, names] of entries) for (const name of names) firstNames.push(name);
+			}
 			const surnames = [...legacySurnameMod.default.names];
-			if (dictionaries?.surnames) for (const names of Object.values(dictionaries.surnames)) for (const name of names) surnames.push(name);
+			if (dictionaries?.surnames) {
+				const entries = languages === void 0 ? Object.entries(dictionaries.surnames) : Object.entries(dictionaries.surnames).filter(([language]) => languages.includes(language));
+				for (const [, names] of entries) for (const name of names) surnames.push(name);
+			}
 			const dedup = (arr) => {
 				const seen = /* @__PURE__ */ new Set();
 				const result = [];
@@ -1909,8 +2060,9 @@ const initNameCorpus = (ctx = defaultContext, dictionaries) => {
 				}
 				return result;
 			};
+			const commonWords = new Set(commonWordsMod.default.words.map((word) => word.toLowerCase()));
 			const dedupFirst = dedup(firstNames);
-			const dedupSurnames = dedup(surnames);
+			const dedupSurnames = dedup(surnames).filter((name) => !commonWords.has(name.toLowerCase()));
 			const titles = titleMod.default.tokens;
 			const exclusions = exclusionMod.default.words;
 			ctx.nameCorpus = {
@@ -2007,7 +2159,31 @@ const segmentWords = (fullText) => {
 };
 /** NAME or SURNAME — both represent corpus-matched tokens */
 const isCorpusMatch = (type) => type === TOKEN_TYPE.NAME || type === TOKEN_TYPE.SURNAME;
-const classifyToken = (word, corpus) => {
+const ALL_CAPS_NAME_LINE_RATIO = .9;
+const ALL_CAPS_NAME_LINE_MIN_LETTERS = 3;
+const ALL_CAPS_NAME_LINE_MAX_TOKENS = 6;
+const isAllCapsLineNameShaped = (fullText, start) => {
+	const lineStart = fullText.lastIndexOf("\n", start - 1) + 1;
+	const lineEndIdx = fullText.indexOf("\n", start);
+	const line = fullText.slice(lineStart, lineEndIdx === -1 ? fullText.length : lineEndIdx);
+	if (/\d/.test(line)) return false;
+	const tokens = line.match(/\p{L}[\p{L}\p{M}'\-]*/gu) ?? [];
+	return tokens.length > 0 && tokens.length <= ALL_CAPS_NAME_LINE_MAX_TOKENS;
+};
+const isAllCapsContextLine = (fullText, start) => {
+	const lineStart = fullText.lastIndexOf("\n", start - 1) + 1;
+	const lineEndIdx = fullText.indexOf("\n", start);
+	const line = fullText.slice(lineStart, lineEndIdx === -1 ? fullText.length : lineEndIdx);
+	let letters = 0;
+	let upper = 0;
+	for (const ch of line) if (/\p{L}/u.test(ch)) {
+		letters += 1;
+		if (ch === ch.toUpperCase() && ch !== ch.toLowerCase()) upper += 1;
+	}
+	if (letters < ALL_CAPS_NAME_LINE_MIN_LETTERS) return false;
+	return upper / letters >= ALL_CAPS_NAME_LINE_RATIO;
+};
+const classifyToken = (word, corpus, fullText) => {
 	const { text, start, end } = word;
 	const lower = text.toLowerCase();
 	const stripped = text.endsWith(".") ? text.slice(0, -1).toLowerCase() : lower;
@@ -2023,6 +2199,26 @@ const classifyToken = (word, corpus) => {
 		start,
 		end
 	};
+	if (text.length === 1 && UPPER_START_RE.test(text) && fullText[end] === ".") {
+		const lineStart = fullText.lastIndexOf("\n", start - 1) + 1;
+		const before = fullText.slice(lineStart, start).trimEnd();
+		const lastWord = /\p{L}[\p{L}\p{M}'\-]*$/u.exec(before)?.[0];
+		if (lastWord) {
+			const lookup = (token) => isFirstNameToken(token, corpus) || isFirstNameToken((token[0] ?? "") + token.slice(1).toLowerCase(), corpus);
+			if (lookup(lastWord)) return {
+				text,
+				type: TOKEN_TYPE.ABBREVIATION,
+				start,
+				end
+			};
+		}
+		return {
+			text,
+			type: TOKEN_TYPE.OTHER,
+			start,
+			end
+		};
+	}
 	if (corpus.excludedWords.has(lower)) return {
 		text,
 		type: TOKEN_TYPE.OTHER,
@@ -2035,12 +2231,29 @@ const classifyToken = (word, corpus) => {
 		start,
 		end
 	};
-	if (text.length > 3 && ALL_UPPER_RE.test(text)) return {
-		text,
-		type: TOKEN_TYPE.OTHER,
-		start,
-		end
-	};
+	if (text.length >= 3 && ALL_UPPER_RE.test(text)) {
+		if (isAllCapsContextLine(fullText, start) && isAllCapsLineNameShaped(fullText, start)) {
+			const titleCased = (text[0] ?? "") + text.slice(1).toLowerCase();
+			if (isFirstNameToken(titleCased, corpus)) return {
+				text,
+				type: TOKEN_TYPE.NAME,
+				start,
+				end
+			};
+			if (isSurnameToken(titleCased, corpus)) return {
+				text,
+				type: TOKEN_TYPE.SURNAME,
+				start,
+				end
+			};
+		}
+		return {
+			text,
+			type: TOKEN_TYPE.OTHER,
+			start,
+			end
+		};
+	}
 	if (!UPPER_START_RE.test(text)) return {
 		text,
 		type: TOKEN_TYPE.OTHER,
@@ -2085,7 +2298,7 @@ const classifyToken = (word, corpus) => {
 const detectNameCorpus = (fullText, ctx = defaultContext) => {
 	const corpus = getCorpus(ctx);
 	if (!corpus) return [];
-	const tokens = segmentWords(fullText).map((w) => classifyToken(w, corpus));
+	const tokens = segmentWords(fullText).map((w) => classifyToken(w, corpus, fullText));
 	const entities = [];
 	const consumed = /* @__PURE__ */ new Set();
 	for (let i = 0; i < tokens.length; i++) {
@@ -2123,6 +2336,8 @@ const detectNameCorpus = (fullText, ctx = defaultContext) => {
 		else if (hasAbbreviation && hasCorpusName) score = .7;
 		else if (hasFirstName && chain.length === 1) {
 			if (isSentenceStart(fullText, token.start)) continue;
+			const first = chain[0];
+			if (first && ALL_UPPER_RE.test(first.text) && first.text.length >= 3) continue;
 			score = .5;
 		} else if (!hasFirstName && chain.length === 1 && chain[0]?.type === TOKEN_TYPE.SURNAME) continue;
 		else if (hasTitle && chain.length === 1) continue;
@@ -2149,6 +2364,115 @@ const detectNameCorpus = (fullText, ctx = defaultContext) => {
 	return entities;
 };
 //#endregion
+//#region src/detectors/signatures.ts
+const SLASH_S_RE = /\/s\/[ \t]*/g;
+const LABELLED_NAME_RE = /\b(?:By|Name)[ \t]*:[ \t]*(?:\/s\/[ \t]+)?([^\n]*?)(?=\s{3,}|[\t]|\n|$)/gi;
+const WITNESS_ANCHOR_RE = /\bIN WITNESS WHEREOF\b[^\n]*\n/gi;
+const NAME_PARTICLE = "(?:de|del|della|der|den|di|du|el|la|le|van|von|y|zu|af|ben|bin|al|d'|d’)";
+const CAP_TOKEN = "\\p{Lu}[\\p{L}\\p{M}.'\\-]{0,30}";
+const NAME_SHAPE_RE = new RegExp(`^${CAP_TOKEN}(?:[ \\t]+(?:${NAME_PARTICLE}|${CAP_TOKEN})){1,4}$`, "u");
+const MAX_NAME_LEN = 60;
+const POST_NOMINAL_SUFFIX_RE = /,\s*(?:Jr|Sr|II|III|IV|V|Esq|Esquire|M\.?D|Ph\.?D|J\.?D|LL\.?M|MBA|CPA|PE|RN|DDS|DVM|DO|MD|CFA|CFP)\.?\s*$/i;
+const COLUMN_SEPARATOR_RE = /\s{3,}|\t+/;
+const ORG_SUFFIX_RE = /\b(?:INC\.?|LLC|LLP|LP|CORP\.?|CORPORATION|LTD\.?|GMBH|AG|SE|KG|OHG|SA|SAS|SARL|S\.A\.?|S\.P\.A\.?|PLC|N\.A\.?|N\.V\.?|B\.V\.?|PTY\s+LTD\.?|CO\.|S\.R\.O\.?|A\.S\.?|Z\.S\.?|S\.\s*P\.?|LTDA\.?|EIRELI|EPP|S\/A)\b/i;
+const IMAGE_STUB_RE = /^(?:\[?img|\[image|\[logo|\(logo\))/i;
+const normaliseCandidate = (text) => {
+	let candidate = text.trim();
+	candidate = candidate.replace(POST_NOMINAL_SUFFIX_RE, "").trim();
+	return candidate.split(COLUMN_SEPARATOR_RE)[0]?.trim() ?? candidate;
+};
+const isNameShape = (text) => {
+	if (text.length === 0 || text.length > MAX_NAME_LEN) return false;
+	if (text.length < 3) return false;
+	if (!NAME_SHAPE_RE.test(text)) return false;
+	return true;
+};
+const findLineEnd = (text, pos) => {
+	const idx = text.indexOf("\n", pos);
+	return idx === -1 ? text.length : idx;
+};
+const tryEmit = (results, fullText, start, end, score) => {
+	const raw = fullText.slice(start, end);
+	if (ORG_SUFFIX_RE.test(raw)) return false;
+	const candidate = normaliseCandidate(raw);
+	if (!isNameShape(candidate)) return false;
+	const offset = raw.indexOf(candidate);
+	if (offset < 0) return false;
+	const absStart = start + offset;
+	results.push({
+		start: absStart,
+		end: absStart + candidate.length,
+		label: "person",
+		text: candidate,
+		score,
+		source: DETECTION_SOURCES.TRIGGER
+	});
+	return true;
+};
+const tryEmitForwardLines = (results, fullText, fromPos, maxLines, score) => {
+	let pos = fromPos;
+	for (let i = 0; i < maxLines; i++) {
+		if (pos >= fullText.length) return false;
+		const lineEnd = findLineEnd(fullText, pos);
+		const line = fullText.slice(pos, lineEnd).trim();
+		if (line.length > 0 && !IMAGE_STUB_RE.test(line)) {
+			if (tryEmit(results, fullText, pos, lineEnd, score)) return true;
+		}
+		pos = lineEnd + 1;
+	}
+	return false;
+};
+const findPrevLine = (fullText, pos) => {
+	let cursor = pos - 1;
+	while (cursor >= 0 && fullText.charAt(cursor) !== "\n") cursor--;
+	while (cursor >= 0) {
+		let lineStart = cursor;
+		while (lineStart > 0 && fullText.charAt(lineStart - 1) !== "\n") lineStart -= 1;
+		const lineEnd = cursor;
+		const line = fullText.slice(lineStart, lineEnd).trim();
+		if (line.length > 0 && !IMAGE_STUB_RE.test(line)) return {
+			lineStart,
+			lineEnd
+		};
+		cursor = lineStart - 1;
+	}
+	return null;
+};
+const detectSignatures = (fullText, _ctx = defaultContext) => {
+	const results = [];
+	SLASH_S_RE.lastIndex = 0;
+	for (let m = SLASH_S_RE.exec(fullText); m !== null; m = SLASH_S_RE.exec(fullText)) {
+		const afterMark = m.index + m[0].length;
+		const lineEnd = findLineEnd(fullText, afterMark);
+		if (fullText.slice(afterMark, lineEnd).trim().length > 0) {
+			const firstCell = fullText.slice(afterMark, lineEnd).split(COLUMN_SEPARATOR_RE)[0] ?? "";
+			if (firstCell.trim().length > 0) tryEmit(results, fullText, afterMark, afterMark + firstCell.length, .95);
+		} else tryEmitForwardLines(results, fullText, lineEnd + 1, 4, .9);
+		const prev = findPrevLine(fullText, m.index);
+		if (prev) tryEmit(results, fullText, prev.lineStart, prev.lineEnd, .85);
+	}
+	LABELLED_NAME_RE.lastIndex = 0;
+	for (let m = LABELLED_NAME_RE.exec(fullText); m !== null; m = LABELLED_NAME_RE.exec(fullText)) {
+		const value = m[1];
+		if (value === void 0) continue;
+		const valueStart = m.index + m[0].length - value.length;
+		const valueEnd = valueStart + value.length;
+		if (value.trim().length === 0) {
+			tryEmitForwardLines(results, fullText, valueEnd + 1, 3, .9);
+			continue;
+		}
+		tryEmit(results, fullText, valueStart, valueEnd, .95);
+	}
+	WITNESS_ANCHOR_RE.lastIndex = 0;
+	for (let m = WITNESS_ANCHOR_RE.exec(fullText); m !== null; m = WITNESS_ANCHOR_RE.exec(fullText)) {
+		const search = fullText.slice(m.index, m.index + 600);
+		const sentenceEnd = /[.:;]\s*\n|\n\s*\n/.exec(search);
+		if (!sentenceEnd) continue;
+		tryEmitForwardLines(results, fullText, m.index + sentenceEnd.index + sentenceEnd[0].length, 6, .85);
+	}
+	return results;
+};
+//#endregion
 //#region src/config/titles.ts
 /**
 * Academic and professional title prefixes.
@@ -2391,9 +2715,14 @@ const STDNUM_ENTRIES = [
 	toEntry(at.uid, "tax identification number", .95),
 	toEntry(at.tin, "tax identification number", .9),
 	toEntry(at.businessid, "registration number", .95),
+	toEntry(ch.uid, "registration number", .95),
+	toEntry(au.abn, "tax identification number", .9),
+	toEntry(au.acn, "registration number", .9),
 	toEntry(be.vat, "tax identification number", .95),
 	toEntry(be.nn, "national identification number", .9),
 	toEntry(nl.vat, "tax identification number", .95),
+	toEntry(no.orgnr, "registration number", .9),
+	toEntry(no.mva, "tax identification number", .95),
 	toEntry(dk.vat, "tax identification number", .95),
 	toEntry(dk.cpr, "national identification number", .9),
 	toEntry(fi.vat, "tax identification number", .95),
@@ -2423,6 +2752,7 @@ const STDNUM_ENTRIES = [
 	toEntry(cy.vat, "tax identification number", .95),
 	toEntry(mt.vat, "tax identification number", .95),
 	toEntry(lu.vat, "tax identification number", .95),
+	toEntry(us.ein, "tax identification number", .9),
 	toEntry(br.cpf, "tax identification number", .95),
 	toEntry(br.cnpj, "tax identification number", .95)
 ].filter((e) => e !== null);
@@ -2535,6 +2865,30 @@ const ES_CIF = {
 	score: .95,
 	validator: es.cif
 };
+const AU_ABN_FORMATTED = {
+	pattern: `\\b\\d{2}[^\\S\\n]\\d{3}[^\\S\\n]\\d{3}[^\\S\\n]\\d{3}\\b`,
+	label: "tax identification number",
+	score: .95,
+	validator: au.abn
+};
+const NO_ORGNR_FORMATTED = {
+	pattern: `\\b\\d{3}[^\\S\\n]\\d{3}[^\\S\\n]\\d{3}\\b`,
+	label: "registration number",
+	score: .9,
+	validator: no.orgnr
+};
+const NO_MVA_FORMATTED = {
+	pattern: "\\bNO[^\\S\\n]?\\d{3}[^\\S\\n]?\\d{3}[^\\S\\n]?\\d{3}[^\\S\\n]?MVA\\b",
+	label: "tax identification number",
+	score: .95,
+	validator: no.mva
+};
+const US_EIN_FORMATTED = {
+	pattern: `\\b\\d{2}${DASH}\\d{7}\\b`,
+	label: "tax identification number",
+	score: .95,
+	validator: us.ein
+};
 const BR_CPF_FORMATTED = {
 	pattern: `\\b\\d{3}\\.\\d{3}\\.\\d{3}${DASH}\\d{2}\\b`,
 	label: "tax identification number",
@@ -2655,6 +3009,10 @@ const ALL_REGEX_DEFS = [
 	ES_DNI,
 	ES_NIE,
 	ES_CIF,
+	AU_ABN_FORMATTED,
+	NO_ORGNR_FORMATTED,
+	NO_MVA_FORMATTED,
+	US_EIN_FORMATTED,
 	BR_CPF_FORMATTED,
 	BR_CNPJ_FORMATTED,
 	BR_RG_WITH_SSP,
@@ -3934,6 +4292,47 @@ const MAX_ENTITY_LENGTH = {
 	organization: 80,
 	person: 60
 };
+const MAX_ENTITY_WORDS = { organization: 8 };
+const OPEN_ENDED_SOURCES = new Set(["trigger", "coreference"]);
+const WORD_TOKEN_RE = /\p{L}[\p{L}\p{M}\p{N}'’\-./]*/gu;
+const countWordTokens = (text) => {
+	let count = 0;
+	WORD_TOKEN_RE.lastIndex = 0;
+	while (WORD_TOKEN_RE.exec(text) !== null) count += 1;
+	return count;
+};
+const ALL_CAPS_LINE_LETTER_THRESHOLD = 5;
+const ALL_CAPS_LINE_RATIO = .95;
+const ALL_CAPS_LINE_PROSE_EXTRA_LETTERS = 20;
+const ALL_CAPS_LINE_HEADING_WORD_LIMIT = 5;
+const SECTION_HEADING_PREFIX_RE = /^\s*(?:§\s*)?\d{1,3}(?:\.\d{1,3}){0,4}\.?\s*\p{Lu}/u;
+const LINE_LETTER_RE = /\p{L}/gu;
+const ENTITY_WORD_TOKEN_RE = /\p{L}[\p{L}\p{M}\p{N}'’\-]*/gu;
+const isAllCapsBoilerplateLine = (fullText, start, length) => {
+	const lineStart = fullText.lastIndexOf("\n", start) + 1;
+	const lineEndIdx = fullText.indexOf("\n", start + length);
+	const line = fullText.slice(lineStart, lineEndIdx === -1 ? fullText.length : lineEndIdx);
+	let letterCount = 0;
+	let upperCount = 0;
+	let outsideEntityLetters = 0;
+	const entityRelStart = start - lineStart;
+	const entityRelEnd = entityRelStart + length;
+	LINE_LETTER_RE.lastIndex = 0;
+	for (let m = LINE_LETTER_RE.exec(line); m !== null; m = LINE_LETTER_RE.exec(line)) {
+		const ch = m[0];
+		letterCount += 1;
+		if (ch === ch.toUpperCase() && ch !== ch.toLowerCase()) upperCount += 1;
+		if (m.index < entityRelStart || m.index >= entityRelEnd) outsideEntityLetters += 1;
+	}
+	if (letterCount <= ALL_CAPS_LINE_LETTER_THRESHOLD) return false;
+	if (upperCount / letterCount < ALL_CAPS_LINE_RATIO) return false;
+	if (SECTION_HEADING_PREFIX_RE.test(line)) return true;
+	if (outsideEntityLetters >= ALL_CAPS_LINE_PROSE_EXTRA_LETTERS) return true;
+	const entityText = fullText.slice(start, start + length);
+	if ((entityText.match(ENTITY_WORD_TOKEN_RE) ?? []).length > ALL_CAPS_LINE_HEADING_WORD_LIMIT && !entityText.includes(",")) return true;
+	return false;
+};
+const isAllCapsCandidate = (text) => text === text.toUpperCase() && /\p{Lu}/u.test(text);
 const SECTION_NUMBER_RE = /^(?:§\s*)?\d{1,3}(?:\.\d{1,3}){0,4}\.?$/;
 const STANDALONE_YEAR_RE = /^(?:19|20)\d{2}$/;
 const NUMBER_ABBREV_RE = /(?:^|[\s(])(?:č|čís|nr|no|n)\.\s*$/i;
@@ -4073,6 +4472,9 @@ const filterFalsePositives = (entities, ctx = defaultContext, fullText) => {
 		if (TEMPLATE_PLACEHOLDER_RE.test(trimmed)) continue;
 		const maxLen = MAX_ENTITY_LENGTH[normalized.label];
 		if (maxLen && trimmed.length > maxLen && normalized.source !== "legal-form") continue;
+		const maxWords = MAX_ENTITY_WORDS[normalized.label];
+		if (maxWords && OPEN_ENDED_SOURCES.has(normalized.source) && countWordTokens(trimmed) > maxWords) continue;
+		if (fullText && normalized.label === "organization" && isAllCapsCandidate(trimmed) && isAllCapsBoilerplateLine(fullText, normalized.start, trimmed.length)) continue;
 		if (SECTION_NUMBER_RE.test(trimmed) && normalized.source !== "trigger") continue;
 		if (STANDALONE_YEAR_RE.test(trimmed) && normalized.source !== "trigger") continue;
 		if (fullText && normalized.source !== "trigger" && /^\d/.test(trimmed) && NUMBER_ABBREV_RE.test(fullText.slice(Math.max(0, normalized.start - 10), normalized.start))) continue;
@@ -4150,6 +4552,25 @@ const loadAllowList = (ctx) => {
 const EMPTY_ALLOW_LIST = /* @__PURE__ */ new Set();
 /** Sync accessor — returns empty set before init. */
 const getAllowList = (ctx) => ctx.allowList ?? EMPTY_ALLOW_LIST;
+let commonWordsPromise = null;
+let commonWordsCache = null;
+const loadCommonWords = () => {
+	if (commonWordsCache) return Promise.resolve(commonWordsCache);
+	if (commonWordsPromise) return commonWordsPromise;
+	commonWordsPromise = (async () => {
+		try {
+			const mod = await import("./common-words-en.mjs");
+			const set = new Set((mod.default?.words ?? []).map((word) => word.toLowerCase()));
+			commonWordsCache = set;
+			return set;
+		} catch {
+			const empty = /* @__PURE__ */ new Set();
+			commonWordsCache = empty;
+			return empty;
+		}
+	})();
+	return commonWordsPromise;
+};
 /**
 * Curated dictionary entries that are pure dotted
 * single-letter acronyms (e.g. `S.C.`, `D.N.J.`, `C.E.C.`)
@@ -4381,6 +4802,21 @@ const SENTENCE_STARTER_WORDS = new Set([
 const PERSON_CHAIN_BREAK_RE = /[!?;:]|,/u;
 const WORD_CHAR_RE$1 = /[\p{L}\p{N}]/u;
 const isInitialContinuationGap = (text, gap) => /^\p{Lu}$/u.test(text) && /^\.[^\S\n]{1,2}$/u.test(gap) || /^[^\S\n]{1,2}(?:\p{Lu}\.[^\S\n]{1,2})+$/u.test(gap);
+const getCityEntries = (dictionaries, allowedCountries) => {
+	const byCountry = dictionaries?.citiesByCountry;
+	if (!byCountry) return dictionaries?.cities ?? [];
+	const result = [];
+	const append = (entries) => {
+		if (!entries) return;
+		for (const entry of entries) result.push(entry);
+	};
+	if (allowedCountries === null) {
+		for (const entries of Object.values(byCountry)) append(entries);
+		return result;
+	}
+	for (const country of allowedCountries) append(byCountry[country.toUpperCase()]);
+	return result;
+};
 /**
 * Resolve which dictionaries to load based on country
 * and category filters, then build the deny list data.
@@ -4392,21 +4828,24 @@ const isInitialContinuationGap = (text, gap) => /^\p{Lu}$/u.test(text) && /^\.[^
 * Returns null if no dictionaries are provided.
 */
 const buildDenyList = async (config, ctx = defaultContext) => {
-	await initNameCorpus(ctx, config.dictionaries);
+	await initNameCorpus(ctx, config.dictionaries, config.nameCorpusLanguages);
 	await Promise.all([
 		loadStopwords(ctx),
 		loadAllowList(ctx),
 		loadPersonStopwords(ctx),
 		loadAddressStopwords(ctx),
+		loadCommonWords(),
 		loadStreetTypeRe(),
 		loadGenericRoles(ctx)
 	]);
+	const commonWords = await loadCommonWords();
 	const dictionaries = config.dictionaries;
 	const hasDenyList = dictionaries?.denyList && dictionaries?.denyListMeta;
-	const hasCities = dictionaries?.cities && dictionaries.cities.length > 0;
 	const hasCustomDenyList = config.customDenyList !== void 0 && config.customDenyList.length > 0;
-	if (!hasDenyList && !hasCities && !hasCustomDenyList) return buildNameCorpusOnly(config, ctx);
 	const allowedCountries = resolveCountries(config.denyListRegions, config.denyListCountries);
+	const cityEntries = getCityEntries(dictionaries, allowedCountries);
+	const hasCities = cityEntries.length > 0;
+	if (!hasDenyList && !hasCities && !hasCustomDenyList) return buildNameCorpusOnly(config, ctx);
 	const excluded = config.denyListExcludeCategories;
 	const excludeCategories = excluded ? new Set(excluded) : /* @__PURE__ */ new Set();
 	const patternList = [];
@@ -4417,8 +4856,11 @@ const buildDenyList = async (config, ctx = defaultContext) => {
 	const addDenyListEntry = (entry, label, source = "deny-list") => {
 		const normalized = source === "custom-deny-list" ? normalizeForSearch(entry) : normalizeForSearch(entry).replace(/[|\\]/g, "");
 		if (normalized.length === 0) return;
-		if (source !== "custom-deny-list" && label !== "address" && isShortCuratedNoiseAcronym(normalized)) return;
 		const lower = normalized.toLowerCase();
+		if (source !== "custom-deny-list" && label !== "address") {
+			if (SINGLE_WORD_RE.test(normalized) && commonWords.has(lower)) return;
+			if (isShortCuratedNoiseAcronym(normalized)) return;
+		}
 		const existing = patternIndex.get(lower);
 		if (existing !== void 0) {
 			if (!labelList[existing].includes(label)) labelList[existing].push(label);
@@ -4435,10 +4877,12 @@ const buildDenyList = async (config, ctx = defaultContext) => {
 	if (hasDenyList) {
 		const denyListData = dictionaries.denyList;
 		const metaData = dictionaries.denyListMeta;
+		const useScopedNameCorpus = config.nameCorpusLanguages !== void 0;
 		for (const [id, entries] of Object.entries(denyListData)) {
 			const meta = metaData[id];
 			if (!meta) continue;
 			if (!config.enableNameCorpus && meta.category === "Names") continue;
+			if (useScopedNameCorpus && meta.category === "Names") continue;
 			if (excludeCategories.has(meta.category)) continue;
 			if (allowedCountries !== null && meta.country !== null) {
 				if (!allowedCountries.has(meta.country)) continue;
@@ -4446,7 +4890,7 @@ const buildDenyList = async (config, ctx = defaultContext) => {
 			for (const entry of entries) addDenyListEntry(entry, meta.label);
 		}
 	}
-	if (hasCities && !excludeCategories.has("Places")) for (const entry of dictionaries.cities) addDenyListEntry(entry, "address");
+	if (hasCities && !excludeCategories.has("Places")) for (const entry of cityEntries) addDenyListEntry(entry, "address", "city");
 	if (hasCustomDenyList) for (const entry of config.customDenyList) {
 		addDenyListEntry(entry.value, entry.label, "custom-deny-list");
 		for (const variant of entry.variants ?? []) addDenyListEntry(variant, entry.label, "custom-deny-list");
@@ -4541,8 +4985,8 @@ const customMatchHasValidEdges = (fullText, start, end, pattern) => {
 * the search instance was built on a different context
 * (e.g. cachedSearch).
 */
-const ensureDenyListData = async (ctx = defaultContext, dictionaries) => {
-	await initNameCorpus(ctx, dictionaries);
+const ensureDenyListData = async (ctx = defaultContext, dictionaries, nameCorpusLanguages) => {
+	await initNameCorpus(ctx, dictionaries, nameCorpusLanguages);
 	await Promise.all([
 		loadStopwords(ctx),
 		loadAllowList(ctx),
@@ -4721,7 +5165,7 @@ const extendCityDistricts = (entities, fullText) => {
 			entity.text = fullText.slice(entity.start, entity.end);
 		}
 		const afterDistrict = fullText.slice(entity.end);
-		const dashDistrictM = /^[\s]*[-–][\s]*(\p{Lu}\p{Ll}+)/u.exec(afterDistrict);
+		const dashDistrictM = /^[ \t]{1,4}[-–][ \t]*(\p{Lu}\p{Ll}+)/u.exec(afterDistrict);
 		if (dashDistrictM && !dashDistrictM[0].includes("\n")) {
 			entity.end += dashDistrictM[0].length;
 			entity.text = fullText.slice(entity.start, entity.end);
@@ -4962,6 +5406,7 @@ const hasHouseNumberNearStreetWord = (fullText, seed) => {
 	const after = fullText.slice(seed.end, Math.min(fullText.length, seed.end + 24));
 	return HOUSE_NUMBER_AFTER_STREET_RE.test(after);
 };
+const isLowercaseStreetWordInProse = (fullText, seed) => /^\p{Ll}/u.test(seed.text) && /^\s+\p{Ll}/u.test(fullText.slice(seed.end, seed.end + 16)) && !hasHouseNumberNearStreetWord(fullText, seed);
 const getUsZipPlusFourContext = (fullText, start, seeds) => {
 	const stateSeed = getUsStateSeedBeforeZip(fullText, start);
 	if (stateSeed !== null) return {
@@ -5022,12 +5467,14 @@ const collectSeeds = (allMatches, sliceStart, sliceEnd, fullText, existingEntiti
 	for (const match of allMatches) {
 		const idx = match.pattern;
 		if (idx < sliceStart || idx >= sliceEnd) continue;
-		seeds.push({
+		const seed = {
 			type: "street-word",
 			start: match.start,
 			end: match.end,
 			text: match.text
-		});
+		};
+		if (isLowercaseStreetWordInProse(fullText, seed)) continue;
+		seeds.push(seed);
 	}
 	for (const e of existingEntities) {
 		if (e.label !== "address") continue;
@@ -5166,6 +5613,7 @@ const NON_ADDRESS_LABELS = new Set([
 ]);
 const expandCluster = async (fullText, cluster, existingEntities) => {
 	const { start, end } = cluster;
+	const seedTypes = new Set(cluster.seeds.map((seed) => seed.type));
 	let leftBound = 0;
 	for (const e of existingEntities) if (NON_ADDRESS_LABELS.has(e.label) && e.end <= start && e.end > leftBound) leftBound = e.end;
 	let leftPos = start;
@@ -5180,6 +5628,10 @@ const expandCluster = async (fullText, cluster, existingEntities) => {
 		if (fullText.slice(p + 1, leftPos).includes("\n")) break;
 		leftPos = p + 1;
 	}
+	if (!(seedTypes.has("street-word") || seedTypes.has("house-number") || seedTypes.has("postal-code") || seedTypes.has("address-trigger"))) return {
+		start: Math.min(leftPos, start),
+		end
+	};
 	let rightPos = end;
 	const remaining = fullText.slice(rightPos);
 	let nearestBoundary = Math.min(remaining.length, 200);
@@ -6088,7 +6540,7 @@ const mergeAdjacent = (entities, fullText) => {
 				break;
 			}
 		}
-		if (!hasLockedBoundary$1(prev) && !(isLegalFormOrganization(prev) && isLegalFormOrganization(entity) && gap.includes(",")) && !gapOccupied && gap.length <= MAX_GAP && GAP_PATTERN.test(gap)) {
+		if (!hasLockedBoundary$1(prev) && !((isLegalFormOrganization(prev) || isLegalFormOrganization(entity)) && gap.includes(",")) && !gapOccupied && gap.length <= MAX_GAP && GAP_PATTERN.test(gap)) {
 			prev.end = entity.end;
 			prev.text = fullText.slice(prev.start, prev.end);
 			prev.score = Math.max(prev.score, entity.score);
@@ -6246,9 +6698,12 @@ const enforceBoundaryConsistency = (entities, fullText) => {
 //#region src/build-unified-search.ts
 const DEFAULT_CUSTOM_REGEX_SCORE$1 = .9;
 const ALNUM_RE = /[\p{L}\p{N}]/u;
+const createAllowedLabelSet$1 = (labels) => labels.length > 0 ? new Set(labels) : null;
+const labelIsAllowed$1 = (label, allowedLabels) => allowedLabels === null || allowedLabels.has(label);
 const buildUnifiedSearch = async (config, gazetteerEntries = [], ctx = defaultContext) => {
 	const legalFormsEnabled = isLegalFormsEnabled(config);
-	const customRegexes = config.enableRegex ? config.customRegexes ?? [] : [];
+	const allowedLabels = createAllowedLabelSet$1(config.enableHotwordRules === true ? expandLabelsForHotwordRules(config.labels) : config.labels);
+	const customRegexes = config.enableRegex ? (config.customRegexes ?? []).filter((entry) => labelIsAllowed$1(entry.label, allowedLabels)) : [];
 	const [legalForms, triggers, denyListData, streetTypes, currencyPatterns, datePatterns, signingPatterns] = await Promise.all([
 		legalFormsEnabled ? buildLegalFormPatterns() : Promise.resolve([]),
 		config.enableTriggerPhrases ? buildTriggerPatterns() : Promise.resolve({
@@ -6257,22 +6712,30 @@ const buildUnifiedSearch = async (config, gazetteerEntries = [], ctx = defaultCo
 		}),
 		config.enableDenyList ? buildDenyList(config, ctx) : Promise.resolve(null),
 		buildStreetTypePatterns(),
-		getCurrencyPatterns(),
-		getDatePatterns(),
-		getSigningClausePatterns()
+		config.enableRegex && labelIsAllowed$1("monetary amount", allowedLabels) ? getCurrencyPatterns() : Promise.resolve([]),
+		config.enableRegex && labelIsAllowed$1("date", allowedLabels) ? getDatePatterns() : Promise.resolve([]),
+		config.enableRegex && labelIsAllowed$1("address", allowedLabels) ? getSigningClausePatterns() : Promise.resolve([])
 	]);
-	const allRegex = [
-		...REGEX_PATTERNS,
-		...currencyPatterns,
-		...datePatterns,
-		...signingPatterns
-	];
-	const regexMeta = [
-		...REGEX_META,
-		...currencyPatterns.map(() => CURRENCY_PATTERN_META),
-		...datePatterns.map(() => DATE_PATTERN_META),
-		...signingPatterns.map(() => SIGNING_CLAUSE_META)
-	];
+	const allRegex = [];
+	const regexMeta = [];
+	if (config.enableRegex) for (const [index, pattern] of REGEX_PATTERNS.entries()) {
+		const meta = REGEX_META[index];
+		if (!meta || !labelIsAllowed$1(meta.label, allowedLabels)) continue;
+		allRegex.push(pattern);
+		regexMeta.push(meta);
+	}
+	for (const pattern of currencyPatterns) {
+		allRegex.push(pattern);
+		regexMeta.push(CURRENCY_PATTERN_META);
+	}
+	for (const pattern of datePatterns) {
+		allRegex.push(pattern);
+		regexMeta.push(DATE_PATTERN_META);
+	}
+	for (const pattern of signingPatterns) {
+		allRegex.push(pattern);
+		regexMeta.push(SIGNING_CLAUSE_META);
+	}
 	const customRegexMeta = customRegexes.map((entry) => ({
 		label: entry.label,
 		score: entry.score ?? DEFAULT_CUSTOM_REGEX_SCORE$1,
@@ -6764,7 +7227,7 @@ const configKey = (config, gazetteerEntries) => {
 		score: entry.score ?? DEFAULT_CUSTOM_REGEX_SCORE
 	})).sort().join("\n") : "";
 	const gazFingerprint = config.enableGazetteer && gazetteerEntries.length > 0 ? gazetteerEntries.map((e) => `${e.id}:${e.canonical}:${e.label}:${[...e.variants].sort().join(",")}`).toSorted().join(";") : "";
-	return `${config.enableDenyList}:${config.enableTriggerPhrases}:${legalFormsEnabled}:${config.enableNameCorpus}:${config.enableRegex}:${config.denyListCountries?.toSorted().join(",") ?? ""}:${config.denyListRegions?.toSorted().join(",") ?? ""}:${config.denyListExcludeCategories?.toSorted().join(",") ?? ""}:${customDenyFingerprint}:${customRegexFingerprint}:${config.enableGazetteer}:${gazFingerprint}`;
+	return `${config.enableDenyList}:${config.enableTriggerPhrases}:${legalFormsEnabled}:${config.enableNameCorpus}:${config.nameCorpusLanguages?.toSorted().join(",") ?? ""}:${config.enableRegex}:${config.labels.toSorted().join(",")}:${config.denyListCountries?.toSorted().join(",") ?? ""}:${config.denyListRegions?.toSorted().join(",") ?? ""}:${config.denyListExcludeCategories?.toSorted().join(",") ?? ""}:${customDenyFingerprint}:${customRegexFingerprint}:${config.enableGazetteer}:${gazFingerprint}`;
 };
 /**
 * Get or build a cached search instance. Cache state
@@ -6784,6 +7247,13 @@ const getCachedSearch = async (config, gazetteerEntries, ctx) => {
 	return result;
 };
 /**
+* Pre-build and cache the unified search instance for a
+* pipeline configuration. Use the same context in
+* `runPipeline` to reuse the prepared automata without
+* passing `cachedSearch` around manually.
+*/
+const preparePipelineSearch = ({ config, gazetteerEntries = [], context }) => getCachedSearch(config, gazetteerEntries, context ?? defaultContext);
+/**
 * Run the full detection pipeline.
 *
 * Two TextSearch instances scan the text (regex +
@@ -6840,7 +7310,7 @@ const runPipeline = async (options) => {
 		warmAddressStopKeywords(),
 		hotwordInit
 	]);
-	if (cachedSearch && config.enableDenyList) await ensureDenyListData(ctx, config.dictionaries);
+	if (cachedSearch && config.enableDenyList) await ensureDenyListData(ctx, config.dictionaries, config.nameCorpusLanguages);
 	let zones = [];
 	if (config.enableZoneClassification && zoneInitOk) {
 		zones = classifyZones(fullText, ctx);
@@ -6849,7 +7319,11 @@ const runPipeline = async (options) => {
 	checkAbort(signal);
 	const hotwordsActive = enableHotwords && hotwordInitOk;
 	const preHotwordAllowedLabels = hotwordsActive ? createAllowedLabelSetFromLabels(expandLabelsForHotwordRules(config.labels)) : allowedLabels;
-	const search = cachedSearch ?? await getCachedSearch(config, gazetteerEntries, ctx);
+	const searchConfig = hotwordsActive ? {
+		...config,
+		labels: [...expandLabelsForHotwordRules(config.labels)]
+	} : config;
+	const search = cachedSearch ?? await getCachedSearch(searchConfig, gazetteerEntries, ctx);
 	checkAbort(signal);
 	const { regexMatches, customRegexMatches, literalMatches } = runUnifiedSearch(search, fullText);
 	const { slices } = search;
@@ -6865,11 +7339,13 @@ const runPipeline = async (options) => {
 	const rawTriggerEntities = config.enableTriggerPhrases ? processTriggerMatches(regexMatches, slices.triggers.start, slices.triggers.end, fullText, search.triggerRules) : [];
 	const triggerEntities = filterAllowedLabels(rawTriggerEntities, preHotwordAllowedLabels);
 	if (triggerEntities.length > 0) log("trigger-phrases", `${triggerEntities.length} matches`);
+	const signatureEntities = filterAllowedLabels(detectSignatures(fullText, ctx), preHotwordAllowedLabels);
+	if (signatureEntities.length > 0) log("signatures", `${signatureEntities.length} matches`);
 	checkAbort(signal);
 	let rawNameCorpusEntities = [];
 	let nameCorpusEntities = [];
 	if (config.enableNameCorpus && !config.enableDenyList) {
-		await initNameCorpus(ctx, config.dictionaries);
+		await initNameCorpus(ctx, config.dictionaries, config.nameCorpusLanguages);
 		checkAbort(signal);
 		rawNameCorpusEntities = detectNameCorpus(fullText, ctx);
 		nameCorpusEntities = filterAllowedLabels(rawNameCorpusEntities, preHotwordAllowedLabels);
@@ -6921,6 +7397,7 @@ const runPipeline = async (options) => {
 	checkAbort(signal);
 	const preAddressEntities = [
 		...triggerEntities,
+		...signatureEntities,
 		...regexEntities,
 		...legalFormEntities,
 		...nameCorpusEntities,
@@ -7568,6 +8045,6 @@ const levenshtein = (rawA, rawB) => {
 //#region src/wasm.ts
 initTextSearch(TextSearch);
 //#endregion
-export { CURRENCY_PATTERN_META, DATE_PATTERN_META, DEFAULT_ENTITY_LABELS, DEFAULT_OPERATOR_CONFIG, DETECTION_SOURCES, DETECTOR_PRIORITY, OPERATOR_REGISTRY, OPERATOR_TYPES, REGEX_META, REGEX_PATTERNS, REGIONS, ZONE_SCORE_ADJUSTMENTS, applyHotwordRules, applyZoneAdjustments, boostNearMissEntities, buildDenyList, buildGazetteerPatterns, buildLegalFormPatterns, buildPlaceholderMap, buildStreetTypePatterns, buildTriggerPatterns, buildUnifiedSearch, chunkText, classifyZones, computeChunkOffsets, corefKey, createPipelineContext, deanonymise, decodeSpans, decodeTokenSpans, detectNameCorpus, ensureDenyListData, exportRedactionKey, extractDefinedTerms, filterFalsePositives, findCoreferenceSpans, getCurrencyPatterns, getDatePatterns, initAddressComponents, initHotwordRules, initNameCorpus, initZoneClassifier, levenshtein, mergeAndDedup, mergeChunkEntities, normalizeForSearch, prepareBatch, processAddressSeeds, processDenyListMatches, processGazetteerMatches, processLegalFormMatches, processRegexMatches, processTriggerMatches, propagateOrgNames, redactText, resolveCountries, resolveOperator, runPipeline, runUnifiedSearch, sanitizeEntities, tokenizeText, warmLegalRoleHeads };
+export { CURRENCY_PATTERN_META, DATE_PATTERN_META, DEFAULT_ENTITY_LABELS, DEFAULT_OPERATOR_CONFIG, DETECTION_SOURCES, DETECTOR_PRIORITY, OPERATOR_REGISTRY, OPERATOR_TYPES, REGEX_META, REGEX_PATTERNS, REGIONS, ZONE_SCORE_ADJUSTMENTS, applyHotwordRules, applyZoneAdjustments, boostNearMissEntities, buildDenyList, buildGazetteerPatterns, buildLegalFormPatterns, buildPlaceholderMap, buildStreetTypePatterns, buildTriggerPatterns, buildUnifiedSearch, chunkText, classifyZones, computeChunkOffsets, corefKey, createPipelineContext, deanonymise, decodeSpans, decodeTokenSpans, detectNameCorpus, ensureDenyListData, exportRedactionKey, extractDefinedTerms, filterFalsePositives, findCoreferenceSpans, getCurrencyPatterns, getDatePatterns, initAddressComponents, initHotwordRules, initNameCorpus, initZoneClassifier, levenshtein, mergeAndDedup, mergeChunkEntities, normalizeForSearch, prepareBatch, preparePipelineSearch, processAddressSeeds, processDenyListMatches, processGazetteerMatches, processLegalFormMatches, processRegexMatches, processTriggerMatches, propagateOrgNames, redactText, resolveCountries, resolveOperator, runPipeline, runUnifiedSearch, sanitizeEntities, tokenizeText, warmLegalRoleHeads };
 
 //# sourceMappingURL=wasm.mjs.map