@stll/anonymize-wasm 1.4.3 → 1.4.5

package/dist/common-words-en.mjs

@@ -412,7 +412,6 @@ var words = [
 	"ant",
 	"antarctica",
 	"antenna",
-	"anthony",
 	"anthropology",
 	"anti",
 	"antibodies",

package/dist/constants.mjs

@@ -25,6 +25,7 @@ const DETECTION_SOURCES = {
 	DENY_LIST: "deny-list",
 	LEGAL_FORM: "legal-form",
 	GAZETTEER: "gazetteer",
+	COUNTRY: "country",
 	NER: "ner",
 	COREFERENCE: "coreference"
 };
@@ -39,6 +40,7 @@ const DETECTOR_PRIORITY = {
 	[DETECTION_SOURCES.TRIGGER]: 4,
 	[DETECTION_SOURCES.LEGAL_FORM]: 3,
 	[DETECTION_SOURCES.REGEX]: 3,
+	[DETECTION_SOURCES.COUNTRY]: 3,
 	[DETECTION_SOURCES.DENY_LIST]: 2,
 	[DETECTION_SOURCES.COREFERENCE]: 2,
 	[DETECTION_SOURCES.NER]: 1
@@ -62,6 +64,7 @@ const DEFAULT_ENTITY_LABELS = [
 	"organization",
 	"phone number",
 	"address",
+	"country",
 	"email address",
 	"date",
 	"date of birth",

package/dist/constants.mjs.map

@@ -1 +1 @@
-{"version":3,"file":"constants.mjs","names":[],"sources":["../../src/constants.ts"],"sourcesContent":["/**\n * Runtime-free constants for the anonymization pipeline.\n *\n * This module is the SSR-safe / browser-safe entrypoint:\n * importing it must not pull in `@stll/text-search`,\n * `@stll/anonymize-wasm`, or any other runtime-bearing\n * module. Consumers that only need the static label list,\n * detection-source identifiers, or operator names import\n * from `@stll/anonymize/constants` (or\n * `@stll/anonymize-wasm/constants`) without paying the\n * wasm / regex-set startup cost.\n *\n * `types.ts` re-exports these for back-compat, so existing\n * `import { DEFAULT_ENTITY_LABELS } from \"@stll/anonymize\"`\n * call sites keep working.\n */\n\n/**\n * Source of a detected entity span.\n * Ordered by detection layer in the pipeline.\n */\nexport const DETECTION_SOURCES = {\n  TRIGGER: \"trigger\",\n  REGEX: \"regex\",\n  DENY_LIST: \"deny-list\",\n  LEGAL_FORM: \"legal-form\",\n  GAZETTEER: \"gazetteer\",\n  NER: \"ner\",\n  COREFERENCE: \"coreference\",\n} as const;\n\nexport type DetectionSource =\n  (typeof DETECTION_SOURCES)[keyof typeof DETECTION_SOURCES];\n\n/**\n * Priority levels for detection sources.\n * Higher = more structurally reliable. Used during\n * overlap resolution so deterministic detectors beat\n * probabilistic ones regardless of raw score.\n */\nexport const DETECTOR_PRIORITY = {\n  [DETECTION_SOURCES.GAZETTEER]: 5,\n  [DETECTION_SOURCES.TRIGGER]: 4,\n  [DETECTION_SOURCES.LEGAL_FORM]: 3,\n  [DETECTION_SOURCES.REGEX]: 3,\n  [DETECTION_SOURCES.DENY_LIST]: 2,\n  [DETECTION_SOURCES.COREFERENCE]: 2,\n  [DETECTION_SOURCES.NER]: 1,\n} as const satisfies Record<DetectionSource, number>;\n\n/**\n * Anonymization operator types. Each operator defines\n * how a confirmed entity is replaced in the output.\n */\nexport const OPERATOR_TYPES = [\"replace\", \"redact\"] as const;\n\nexport type OperatorType = (typeof OPERATOR_TYPES)[number];\n\n/**\n * Canonical entity labels used across the pipeline.\n * NER models may use different native labels; the bench\n * NER wrapper maps model output to these canonical names.\n *\n * These labels are ephemeral: entities are regenerated on\n * every pipeline run and never persisted to the database.\n * Renaming a label here requires no migration.\n */\nexport const DEFAULT_ENTITY_LABELS = [\n  \"person\",\n  \"organization\",\n  \"phone number\",\n  \"address\",\n  \"email address\",\n  \"date\",\n  \"date of birth\",\n  \"bank account number\",\n  \"iban\",\n  \"tax identification number\",\n  \"identity card number\",\n  \"birth number\",\n  \"national identification number\",\n  \"social security number\",\n  \"registration number\",\n  \"credit card number\",\n  \"passport number\",\n  \"monetary amount\",\n  \"land parcel\",\n  \"misc\",\n] as const;\n"],"mappings":";;;;;;;;;;;;;;;;;;;;;AAqBA,MAAa,oBAAoB;CAC/B,SAAS;CACT,OAAO;CACP,WAAW;CACX,YAAY;CACZ,WAAW;CACX,KAAK;CACL,aAAa;CACd;;;;;;;AAWD,MAAa,oBAAoB;EAC9B,kBAAkB,YAAY;EAC9B,kBAAkB,UAAU;EAC5B,kBAAkB,aAAa;EAC/B,kBAAkB,QAAQ;EAC1B,kBAAkB,YAAY;EAC9B,kBAAkB,cAAc;EAChC,kBAAkB,MAAM;CAC1B;;;;;AAMD,MAAa,iBAAiB,CAAC,WAAW,SAAS;;;;;;;;;;AAanD,MAAa,wBAAwB;CACnC;CACA;CACA;CACA;CACA;CACA;CACA;CACA;CACA;CACA;CACA;CACA;CACA;CACA;CACA;CACA;CACA;CACA;CACA;CACA;CACD"}
+{"version":3,"file":"constants.mjs","names":[],"sources":["../../src/constants.ts"],"sourcesContent":["/**\n * Runtime-free constants for the anonymization pipeline.\n *\n * This module is the SSR-safe / browser-safe entrypoint:\n * importing it must not pull in `@stll/text-search`,\n * `@stll/anonymize-wasm`, or any other runtime-bearing\n * module. Consumers that only need the static label list,\n * detection-source identifiers, or operator names import\n * from `@stll/anonymize/constants` (or\n * `@stll/anonymize-wasm/constants`) without paying the\n * wasm / regex-set startup cost.\n *\n * `types.ts` re-exports these for back-compat, so existing\n * `import { DEFAULT_ENTITY_LABELS } from \"@stll/anonymize\"`\n * call sites keep working.\n */\n\n/**\n * Source of a detected entity span.\n * Ordered by detection layer in the pipeline.\n */\nexport const DETECTION_SOURCES = {\n  TRIGGER: \"trigger\",\n  REGEX: \"regex\",\n  DENY_LIST: \"deny-list\",\n  LEGAL_FORM: \"legal-form\",\n  GAZETTEER: \"gazetteer\",\n  COUNTRY: \"country\",\n  NER: \"ner\",\n  COREFERENCE: \"coreference\",\n} as const;\n\nexport type DetectionSource =\n  (typeof DETECTION_SOURCES)[keyof typeof DETECTION_SOURCES];\n\n/**\n * Priority levels for detection sources.\n * Higher = more structurally reliable. Used during\n * overlap resolution so deterministic detectors beat\n * probabilistic ones regardless of raw score.\n */\nexport const DETECTOR_PRIORITY = {\n  [DETECTION_SOURCES.GAZETTEER]: 5,\n  [DETECTION_SOURCES.TRIGGER]: 4,\n  [DETECTION_SOURCES.LEGAL_FORM]: 3,\n  [DETECTION_SOURCES.REGEX]: 3,\n  [DETECTION_SOURCES.COUNTRY]: 3,\n  [DETECTION_SOURCES.DENY_LIST]: 2,\n  [DETECTION_SOURCES.COREFERENCE]: 2,\n  [DETECTION_SOURCES.NER]: 1,\n} as const satisfies Record<DetectionSource, number>;\n\n/**\n * Anonymization operator types. Each operator defines\n * how a confirmed entity is replaced in the output.\n */\nexport const OPERATOR_TYPES = [\"replace\", \"redact\"] as const;\n\nexport type OperatorType = (typeof OPERATOR_TYPES)[number];\n\n/**\n * Canonical entity labels used across the pipeline.\n * NER models may use different native labels; the bench\n * NER wrapper maps model output to these canonical names.\n *\n * These labels are ephemeral: entities are regenerated on\n * every pipeline run and never persisted to the database.\n * Renaming a label here requires no migration.\n */\nexport const DEFAULT_ENTITY_LABELS = [\n  \"person\",\n  \"organization\",\n  \"phone number\",\n  \"address\",\n  \"country\",\n  \"email address\",\n  \"date\",\n  \"date of birth\",\n  \"bank account number\",\n  \"iban\",\n  \"tax identification number\",\n  \"identity card number\",\n  \"birth number\",\n  \"national identification number\",\n  \"social security number\",\n  \"registration number\",\n  \"credit card number\",\n  \"passport number\",\n  \"monetary amount\",\n  \"land parcel\",\n  \"misc\",\n] as const;\n"],"mappings":";;;;;;;;;;;;;;;;;;;;;AAqBA,MAAa,oBAAoB;CAC/B,SAAS;CACT,OAAO;CACP,WAAW;CACX,YAAY;CACZ,WAAW;CACX,SAAS;CACT,KAAK;CACL,aAAa;CACd;;;;;;;AAWD,MAAa,oBAAoB;EAC9B,kBAAkB,YAAY;EAC9B,kBAAkB,UAAU;EAC5B,kBAAkB,aAAa;EAC/B,kBAAkB,QAAQ;EAC1B,kBAAkB,UAAU;EAC5B,kBAAkB,YAAY;EAC9B,kBAAkB,cAAc;EAChC,kBAAkB,MAAM;CAC1B;;;;;AAMD,MAAa,iBAAiB,CAAC,WAAW,SAAS;;;;;;;;;;AAanD,MAAa,wBAAwB;CACnC;CACA;CACA;CACA;CACA;CACA;CACA;CACA;CACA;CACA;CACA;CACA;CACA;CACA;CACA;CACA;CACA;CACA;CACA;CACA;CACA;CACD"}

package/dist/constants2.d.mts

@@ -25,6 +25,7 @@ declare const DETECTION_SOURCES: {
   readonly DENY_LIST: "deny-list";
   readonly LEGAL_FORM: "legal-form";
   readonly GAZETTEER: "gazetteer";
+  readonly COUNTRY: "country";
   readonly NER: "ner";
   readonly COREFERENCE: "coreference";
 };
@@ -40,6 +41,7 @@ declare const DETECTOR_PRIORITY: {
   readonly trigger: 4;
   readonly "legal-form": 3;
   readonly regex: 3;
+  readonly country: 3;
   readonly "deny-list": 2;
   readonly coreference: 2;
   readonly ner: 1;
@@ -59,7 +61,7 @@ type OperatorType = (typeof OPERATOR_TYPES)[number];
  * every pipeline run and never persisted to the database.
  * Renaming a label here requires no migration.
  */
-declare const DEFAULT_ENTITY_LABELS: readonly ["person", "organization", "phone number", "address", "email address", "date", "date of birth", "bank account number", "iban", "tax identification number", "identity card number", "birth number", "national identification number", "social security number", "registration number", "credit card number", "passport number", "monetary amount", "land parcel", "misc"];
+declare const DEFAULT_ENTITY_LABELS: readonly ["person", "organization", "phone number", "address", "country", "email address", "date", "date of birth", "bank account number", "iban", "tax identification number", "identity card number", "birth number", "national identification number", "social security number", "registration number", "credit card number", "passport number", "monetary amount", "land parcel", "misc"];
 //#endregion
 export { OPERATOR_TYPES as a, DetectionSource as i, DETECTION_SOURCES as n, OperatorType as o, DETECTOR_PRIORITY as r, DEFAULT_ENTITY_LABELS as t };
 //# sourceMappingURL=constants2.d.mts.map

package/dist/legal-form-leading-clauses.mjs

@@ -6,7 +6,12 @@ var en = {
 		"by and between",
 		"is between"
 	],
-	"directPrefixes": ["by"]
+	"directPrefixes": [
+		"by",
+		"among",
+		"amongst",
+		"between"
+	]
 };
 var legal_form_leading_clauses_default = {
 	_comment,

package/dist/names-exclusions.mjs

@@ -144,16 +144,28 @@ var words = [
 	"subject",
 	"supplier",
 	"tarif",
+	"tato",
+	"tedy",
+	"tehdy",
 	"tel",
 	"telefon",
 	"tenant",
+	"tento",
 	"termín",
 	"that",
 	"the",
 	"therefore",
 	"thereof",
 	"this",
+	"tito",
+	"tomto",
+	"tomuto",
+	"toto",
+	"touto",
 	"tribunal",
+	"tuto",
+	"tyto",
+	"této",
 	"under",
 	"until",
 	"upon",

package/dist/person-stopwords.mjs

@@ -17,6 +17,7 @@ var person_stopwords_default = {
 		"basic",
 		"belgian",
 		"borrower",
+		"branch",
 		"british",
 		"bulgarian",
 		"buyer",
@@ -35,6 +36,8 @@ var person_stopwords_default = {
 		"default",
 		"delaware",
 		"delivery",
+		"department",
+		"division",
 		"dry",
 		"dutch",
 		"english",
@@ -50,6 +53,7 @@ var person_stopwords_default = {
 		"germany",
 		"greek",
 		"guarantor",
+		"headquarters",
 		"hungarian",
 		"indiana",
 		"interest",
@@ -80,6 +84,7 @@ var person_stopwords_default = {
 		"notice",
 		"november",
 		"october",
+		"office",
 		"ohio",
 		"opportunity",
 		"owner",
@@ -102,6 +107,7 @@ var person_stopwords_default = {
 		"section",
 		"seller",
 		"september",
+		"subsidiary",
 		"slovak",
 		"slovenian",
 		"spanish",
@@ -134,6 +140,16 @@ var person_stopwords_default = {
 		"viale",
 		"vico",
 		"vicolo",
+		"tato",
+		"tento",
+		"tito",
+		"tomto",
+		"tomuto",
+		"toto",
+		"touto",
+		"tuto",
+		"tyto",
+		"této",
 		"acheteur",
 		"acheteuse",
 		"acquéreur",

package/dist/wasm.d.mts

@@ -325,6 +325,13 @@ type PipelineConfig = {
    */
   customRegexes?: readonly CustomRegexPattern[];
   enableGazetteer: boolean;
+  /**
+   * Detect country names (ISO 3166-1 names, curated
+   * aliases, alpha-3 codes). Defaults to true. Names
+   * span all manifest languages plus widely-used
+   * additions (Dutch, Russian, Chinese, Arabic, etc.).
+   */
+  enableCountries?: boolean;
   enableNer: boolean;
   enableConfidenceBoost: boolean;
   enableCoreference: boolean;
@@ -387,7 +394,7 @@ declare const CURRENCY_PATTERN_META: Readonly<RegexMeta>;
 declare const processRegexMatches: (allMatches: Match[], sliceStart: number, sliceEnd: number, meta_: readonly RegexMeta[]) => Entity[];
 //#endregion
 //#region src/detectors/deny-list.d.ts
-type DenyListConfig = Pick<PipelineConfig, "enableDenyList" | "enableNameCorpus" | "nameCorpusLanguages" | "denyListCountries" | "denyListRegions" | "denyListExcludeCategories" | "customDenyList" | "dictionaries">;
+type DenyListConfig = Pick<PipelineConfig, "enableDenyList" | "enableNameCorpus" | "nameCorpusLanguages" | "denyListCountries" | "denyListRegions" | "denyListExcludeCategories" | "customDenyList" | "dictionaries" | "enableCountries">;
 /**
  * Source tag for each pattern in the automaton.
  * "deny-list" = standard deny list entry
@@ -452,6 +459,24 @@ declare const ensureDenyListData: (ctx?: PipelineContext, dictionaries?: Diction
  */
 declare const processDenyListMatches: (allMatches: Match[], sliceStart: number, sliceEnd: number, fullText: string, data: DenyListData, ctx?: PipelineContext) => Entity[];
 //#endregion
+//#region src/detectors/countries.d.ts
+/**
+ * Pre-built country patterns + parallel label/source
+ * metadata. Constructed once and reused across pipeline
+ * runs.
+ */
+type CountryData = {
+  /** Maps local pattern index to entity label. Always "country". */labels: string[];
+  /**
+   * Maps local pattern index to the alpha-2 ISO code the
+   * pattern resolves to. Used for downstream coreference /
+   * placeholder grouping.
+   */
+  isoCodes: string[]; /** Maps local pattern index to pattern variant kind. */
+  variants: CountryVariant[];
+};
+type CountryVariant = "name" | "alias" | "alpha3" | "alpha2";
+//#endregion
 //#region src/build-unified-search.d.ts
 type PatternSlice = {
   start: number;
@@ -477,12 +502,14 @@ type UnifiedSearchInstance = {
     denyList: PatternSlice;
     streetTypes: PatternSlice;
     gazetteer: PatternSlice;
+    countries: PatternSlice;
   };
   regexMeta: readonly RegexMeta[];
   customRegexMeta: readonly RegexMeta[];
   triggerRules: readonly TriggerRule[];
   denyListData: DenyListData | null;
   gazetteerData: GazetteerData | null;
+  countryData: CountryData | null;
 };
 declare const buildUnifiedSearch: (config: PipelineConfig, gazetteerEntries?: GazetteerEntry[], ctx?: PipelineContext) => Promise<UnifiedSearchInstance>;
 //#endregion