euredact 0.5.1 → 0.6.0

package/dist/cjs/sdk.js

@@ -86,6 +86,96 @@ function applyReplacements(text, detections, labelFor) {
     parts.push(text.slice(pos));
     return parts.join("");
 }
+/**
+ * Bracket-mask `text` and record where each label landed.
+ *
+ * Returns what is sent to the cloud tier, plus one `WireLabel` per label.
+ * Everything between labels is copied unchanged, so the labels are all it
+ * takes to carry an offset from the masked text back to the original.
+ *
+ * Always `[TYPE]`, whatever label scheme the caller asked for. That is the
+ * syntax the model was trained on, where a placeholder means "already handled,
+ * do not list it". A token or a numbered label would be read as ordinary text,
+ * so the format on the wire is a model contract and not an output option.
+ */
+function maskForCloud(text, detections) {
+    const parts = [];
+    const labels = [];
+    let pos = 0;
+    let out = 0;
+    for (const det of detections) {
+        const start = Math.max(det.start, pos);
+        if (det.end <= start)
+            continue;
+        const label = `[${det.entityType}]`;
+        out += start - pos;
+        parts.push(text.slice(pos, start), label);
+        labels.push([out, out + label.length, start, det.end]);
+        out += label.length;
+        pos = det.end;
+    }
+    parts.push(text.slice(pos));
+    return [parts.join(""), labels];
+}
+/** Index of the last label starting at or before `offset` — strictly before
+ *  when `strict` — or -1. */
+function labelBefore(labels, offset, strict) {
+    let lo = 0;
+    let hi = labels.length;
+    while (lo < hi) {
+        const mid = (lo + hi) >>> 1;
+        if (strict ? labels[mid][0] < offset : labels[mid][0] <= offset)
+            lo = mid + 1;
+        else
+            hi = mid;
+    }
+    return lo - 1;
+}
+/**
+ * Carry the service's spans from the masked text onto the original.
+ *
+ * The service only ever saw `masked`, so its offsets index that. An offset in
+ * copied text shifts by whatever the labels before it added or removed. An
+ * offset inside a label has no counterpart in the original, so it snaps
+ * outward to the edge of the value the label replaced: an address the model
+ * reports around `[POSTAL_CODE]` covers the postal code, never half of it.
+ * Over-masking is the safe direction.
+ *
+ * A span lying wholly inside one label describes the label, not the document.
+ * It is dropped; the local detection behind the label stands.
+ *
+ * Throws `Unplaceable` when a span does not match the text that was sent. Such
+ * a span cannot be placed, and masking where it points would cover the wrong
+ * characters and leave the right ones in the clear.
+ */
+function ontoOriginal(spans, masked, text, labels, Unplaceable) {
+    const placed = [];
+    for (const span of spans) {
+        if (!(0 <= span.start && span.start <= span.end && span.end <= masked.length) ||
+            masked.slice(span.start, span.end) !== span.text) {
+            throw new Unplaceable("span offsets do not match the text that was sent; cannot place the " +
+                "service's detections in the document");
+        }
+        if (span.start === span.end)
+            continue;
+        let start = span.start;
+        let at = labelBefore(labels, span.start, false);
+        if (at >= 0) {
+            const [, maskedEnd, origStart, origEnd] = labels[at];
+            if (span.end <= maskedEnd)
+                continue;
+            start = span.start < maskedEnd ? origStart : origEnd + span.start - maskedEnd;
+        }
+        let end = span.end;
+        at = labelBefore(labels, span.end, true);
+        if (at >= 0) {
+            const [, maskedEnd, , origEnd] = labels[at];
+            end = origEnd + Math.max(0, span.end - maskedEnd);
+        }
+        placed.push({ ...span, start, end, text: text.slice(start, end) });
+    }
+    return placed;
+}
 /**
  * Characters a token suffix is drawn from. No vowels, so a suffix never spells
  * a word; no 0/1/I/O, so it survives being read back by a person; no
@@ -425,7 +515,16 @@ class EuRedact {
         return (det) => `[${det.entityType}]`;
     }
     /**
-     * Send a document to the cloud tier.
+     * Run the rules here, then send what they left to the cloud tier.
+     *
+     * Local-first: the rules engine masks everything it can find on the caller's
+     * machine, and only that masked text is sent. The service looks for what has
+     * no shape to match on — names, employers, diagnoses — and answers with
+     * spans relative to the text it received, which are mapped back onto the
+     * original here.
+     *
+     * This is minimisation, not an exemption: what the rules miss, and the names
+     * and diagnoses the model is there to find, still travel.
      *
      * Options the service cannot honour throw rather than being ignored. Silently
      * dropping one would mean returning a result that does not match what was
@@ -434,6 +533,7 @@ class EuRedact {
      */
     async redactViaCloud(text, options) {
         const { CloudClient } = await Promise.resolve().then(() => __importStar(require("./cloud/client.js")));
+        const { CloudError } = await Promise.resolve().then(() => __importStar(require("./cloud/errors.js")));
         const countries = options.countries ?? null;
         if (!countries || countries.length !== 1) {
             throw new Error('cloud mode needs exactly one country, e.g. { countries: ["BE"] }. The ' +
@@ -451,43 +551,34 @@ class EuRedact {
         if (options.referentialIntegrity) {
             throw new Error("referentialIntegrity is not supported in cloud mode");
         }
-        // detectDates is deliberately NOT forwarded. The service always runs its
-        // rules engine with dates on, because that is what the model was trained
-        // against; the caller's value cannot change that. It is the one ignored
-        // option that is safe to ignore — it can only cause MORE to be detected,
-        // never less, so it cannot produce under-redaction.
         const allowed = this.allowedFor(options);
         const domains = this.domainsFor(options);
-        const result = await new CloudClient().redact(text, { country: countries[0] });
-        if (options.tokenize || allowed.size > 0 || domains.size > 0) {
-            await this.remaskCloudResult(result, text, { tokenize: options.tokenize ?? false, allowed, domains });
-        }
-        return result;
-    }
-    /**
-     * Rebuild the masked text from the service's spans, in place.
-     *
-     * The service masks with bracketed labels only. Any other output option is
-     * applied here, from its spans: the service builds its `redacted_text` from
-     * exactly the spans it returns (entities it reports but cannot place are
-     * listed separately and never applied), so nothing it masked is lost by
-     * masking again from the same spans.
-     *
-     * The service reports offsets in code points and this SDK slices UTF-16
-     * units, so the slice check is what stands between an emoji in the document
-     * and a rebuilt text with the wrong characters masked.
-     */
-    async remaskCloudResult(result, text, opts) {
-        const { CloudError } = await Promise.resolve().then(() => __importStar(require("./cloud/errors.js")));
-        for (const det of result.detections) {
-            if (text.slice(det.start, det.end) !== det.text) {
-                throw new CloudError("span offsets do not match the document; cannot apply tokenize/allowlist locally");
-            }
-        }
-        [result.detections, result.exempted] = applyAllowlist(text, result.detections, opts.allowed, opts.domains);
-        const tokenMapper = opts.tokenize ? new TokenMapper(text, result.detections) : null;
-        result.redactedText = applyReplacements(text, result.detections, this.labelFor(false, tokenMapper));
-        result.tokens = tokenMapper ? tokenMapper.tokens : {};
+        const client = new CloudClient();
+        // Dates are always on, whatever detectDates says: that is the rules output
+        // the model was trained against, and a date of birth the rules can place
+        // has no reason to travel. It is the one ignored option that is safe to
+        // ignore — it can only cause MORE to be detected, never less.
+        //
+        // No allowlist and no tokens on this pass. An exemption says what the
+        // caller wants back, not what may leave; both are applied below, after the
+        // response, to the merged spans.
+        const local = this.redactRules(text, { countries, detectDates: true, cache: options.cache }, new Map(), new Map());
+        const [masked, labels] = maskForCloud(text, local.detections);
+        // The client converts the service's code-point offsets to UTF-16 units of
+        // the text it sent, so everything from here on is in one unit.
+        const remote = await client.redact(masked, { country: countries[0] });
+        // A new array throughout: `local` may be the cached rules result.
+        const [detections, exempted] = applyAllowlist(text, [...local.detections, ...ontoOriginal(remote.detections, masked, text, labels, CloudError)], allowed, domains);
+        detections.sort((a, b) => a.start - b.start || b.end - a.end);
+        const tokenMapper = options.tokenize ? new TokenMapper(text, detections) : null;
+        return {
+            ...local,
+            redactedText: applyReplacements(text, detections, this.labelFor(false, tokenMapper)),
+            detections,
+            source: "cloud",
+            tokens: tokenMapper ? tokenMapper.tokens : {},
+            exempted,
+        };
     }
     redact(text, options = {}) {
         checkCountryArg(options.countries, "countries");
@@ -508,12 +599,16 @@ class EuRedact {
         if (requestedMode !== "rules") {
             throw new Error(`unknown mode ${JSON.stringify(requestedMode)}: expected "rules" or "cloud"`);
         }
+        return this.redactRules(text, options, allowed, domains);
+    }
+    /** The local pipeline: normalise, detect, exempt, mask. */
+    redactRules(text, options, allowed, domains) {
         if (text.length > this.maxInputLength) {
             throw new Error(`Input text length (${text.length.toLocaleString()} chars) exceeds the maximum ` +
                 `(${this.maxInputLength.toLocaleString()} chars). Split the input or ` +
                 `increase maxInputLength when constructing EuRedact.`);
         }
-        const { countries = null, countryHint = null, context = null, chunkOffset = 0, mode = "rules", referentialIntegrity = false, tokenize = false, detectDates = false, } = options;
+        const { countries = null, countryHint = null, context = null, chunkOffset = 0, referentialIntegrity = false, tokenize = false, detectDates = false, } = options;
         // A context makes the result depend on evidence from other chunks, so the
         // text no longer identifies the result. Caching is disabled rather than
         // keyed on the context, whose contents change as chunks arrive.
@@ -531,7 +626,7 @@ class EuRedact {
         // separator, and two different lists must never share a key.
         const allowKey = (allowed.size ? JSON.stringify([...allowed.keys()].sort()) : "")
             + (domains.size ? "|dom=" + JSON.stringify([...domains.keys()].sort()) : "");
-        const cacheMode = `${mode}|dates=${detectDates}|hint=${hintKey}|ri=${referentialIntegrity}|tok=${tokenize}|allow=${allowKey}`;
+        const cacheMode = `rules|dates=${detectDates}|hint=${hintKey}|ri=${referentialIntegrity}|tok=${tokenize}|allow=${allowKey}`;
         let cacheKey;
         if (cache) {
             cacheKey = this.cache.key(normalizedText, countriesTuple, cacheMode);

package/dist/esm/cloud/client.d.ts

@@ -1,11 +1,15 @@
 /**
  * [CLOUD EXTENSION] The cloud client.
  *
- * Speaks to the euRedact inference service: a deterministic rules engine
- * followed by a fine-tuned model asked only *what did the rules miss?* The
- * service returns both the redacted document and located spans, so this
+ * Speaks to the euRedact inference service: a fine-tuned model asked only
+ * *what did the rules miss?* The service returns located spans, so this
  * module's job is transport and translation, not detection.
  *
+ * It sends the text it is given, as given. `redactAsync(text, { mode: "cloud" })`
+ * is what runs the rules engine first and hands this client only the masked
+ * text; call the client directly and whatever you pass is what leaves the
+ * machine. The spans that come back index the text that was sent.
+ *
  * Three things it hides from the caller:
  *
  * - **Whether the document finished inside the service's sync window.** A short
@@ -26,7 +30,6 @@ export interface CloudRedactOptions {
     country: string;
     language?: string;
     priority?: "interactive" | "batch";
-    rulesOnly?: boolean;
     idempotencyKey?: string;
     /** Injectable for tests. Defaults to the platform `fetch`. */
     fetchImpl?: typeof fetch;

package/dist/esm/cloud/client.js

@@ -1,11 +1,15 @@
 /**
  * [CLOUD EXTENSION] The cloud client.
  *
- * Speaks to the euRedact inference service: a deterministic rules engine
- * followed by a fine-tuned model asked only *what did the rules miss?* The
- * service returns both the redacted document and located spans, so this
+ * Speaks to the euRedact inference service: a fine-tuned model asked only
+ * *what did the rules miss?* The service returns located spans, so this
  * module's job is transport and translation, not detection.
  *
+ * It sends the text it is given, as given. `redactAsync(text, { mode: "cloud" })`
+ * is what runs the rules engine first and hands this client only the masked
+ * text; call the client directly and whatever you pass is what leaves the
+ * machine. The spans that come back index the text that was sent.
+ *
  * Three things it hides from the caller:
  *
  * - **Whether the document finished inside the service's sync window.** A short
@@ -153,7 +157,6 @@ export class CloudClient {
             country: options.country.toUpperCase(),
             language: options.language ?? "",
             priority: options.priority ?? "interactive",
-            rules_only: options.rulesOnly ?? false,
         });
         let attempt = 0;
         for (;;) {

package/dist/esm/rules/countries/index.js

@@ -102,18 +102,50 @@ const SECRET_CONTEXT = [
 // ---------------------------------------------------------------------------
 // DOB context keywords (reused across SHARED patterns)
 // ---------------------------------------------------------------------------
+// Every way the documents this engine sees introduce a date of birth.
+//
+// DOB recall was 62.8% -- 33,441 unmasked birth dates -- because this covered
+// seven languages of thirty-one. Eleven countries scored exactly 100% (the
+// listed languages) while twenty sat at 42-51%, which is what proved it was
+// coverage rather than capability (rules-engine#38).
+//
+// Matching is substring, not word-boundary, so every entry was screened against
+// the 29.2-million-character corpus for occurrences inside a longer word. Two
+// were rejected: `fædd` lives inside `fæddur`, and `born` was found inside
+// `gabornagy` and `gabornemeth` -- Hungarian names as e-mail local parts -- so
+// `"born "` with the trailing space is listed instead.
+//
+// DOB_CONTEXT is gone: it was a second, shorter copy missing `nato il`,
+// `nascido`, `geburtstag` and more, so ISO dates were gated on fewer keywords
+// than DD/MM/YYYY ones for no stated reason.
 const DOB_CONTEXT = [
+    // Germanic / Romance -- the original set, kept verbatim.
     "geboren", "geboortedatum", "date de naissance", "né le", "née le",
     "né(e) le", "nee le", "nee(e) le", "date of birth", "DOB",
     "Geburtsdatum", "geboren am", "geboren op", "nascido", "nacido",
     "data di nascita", "nato il", "nata il", "geb.", "geb.datum", "geb ",
     "birth date", "birthday", "naissance", "geboorte", "geburtstag",
-];
-const DOB_ISO_CONTEXT = [
-    "geboren", "geboortedatum", "date de naissance", "né le", "née le",
-    "né(e) le", "nee le", "date of birth", "DOB", "Geburtsdatum",
-    "geboren am", "geboren op", "geb.", "birth date", "birthday",
-    "naissance", "geboorte",
+    // English and Italian forms the list was missing.
+    "born ", "d.o.b", "nato/a il",
+    // Nordic.
+    "født", "fødselsdato", "född", "födelsedatum",
+    "syntynyt", "syntymäaika", "fæddur", "fæðingardagur",
+    // Greek -- Greece and Cyprus.
+    "γεννηθείς", "γεννήθηκε", "ημερομηνία γέννησης",
+    // West Slavic.
+    "urodzony", "urodzona", "data urodzenia",
+    "narozen", "datum narození", "narodený", "dátum narodenia",
+    // Hungarian, Romanian, Bulgarian.
+    "született", "születési", "născut", "născută", "data nașterii",
+    "роден", "родена", "дата на раждане",
+    // South Slavic.
+    "rođen", "rođena", "datum rođenja", "rojen", "rojena", "datum rojstva",
+    // Baltic.
+    "sündinud", "sünniaeg", "dzimis", "dzimusi", "dzimšanas datums",
+    "gimęs", "gimusi", "gimimo data",
+    // Maltese, Irish, Turkish (Cyprus).
+    "twieled", "twieldet", "data tat-twelid",
+    "rugadh", "dáta breithe", "doğum tarihi",
 ];
 const DATE_OF_DEATH_CONTEXT = [
     "overleden", "overlijdensdatum", "date de décès", "décédé le",
@@ -220,7 +252,12 @@ const SHARED = {
         // --- Secret / API Key (connection strings with embedded credentials) ---
         p(EntityType.SECRET, String.raw `(?:mongodb|mysql|postgres(?:ql)?|redis|amqp|rabbitmq):\/\/[^\s:]{1,256}:[^\s@]{1,256}@[^\s]+`, null, "Connection string with credentials"),
         // --- Secret / API Key (assignment-based: KEY=value or KEY: value) ---
-        p(EntityType.SECRET, String.raw `(?<=[:=] )\S{8,}|(?<=[:=])\S{8,}`, "high_entropy", "Assigned secret value", SECRET_CONTEXT, true),
+        // The final character may not be sentence punctuation. `\S` does not stop
+        // at a full stop, so "Reisepass: CA1234567." was claimed as the
+        // ten-character span "CA1234567." -- a passport number plus the sentence's
+        // period, masked as a credential because a "credentials" word sat within
+        // the context window (rules-engine#35).
+        p(EntityType.SECRET, String.raw `(?<=[:=] )\S{7,}[^\s.,;:!?)'"]|(?<=[:=])\S{7,}[^\s.,;:!?)'"]`, "high_entropy", "Assigned secret value", SECRET_CONTEXT, true),
         // --- Secret / API Key (entropy-based fallback for longer tokens) ---
         // Only the *trailing* \b is replaced. It could not match when the token run
         // ended on "-", "+" or "/", so the engine backtracked across two
@@ -238,7 +275,7 @@ const SHARED = {
         p(EntityType.SECRET, String.raw `\b[A-Za-z0-9_\-+/]{24,}[A-Za-z0-9_\-+/=]*(?![A-Za-z0-9_\-+/=])`, "high_entropy", "High-entropy token", SECRET_CONTEXT, true),
         p(EntityType.DOB, String.raw `\b(?:0[1-9]|[12][0-9]|3[01])[/.\-](?:0[1-9]|1[0-2])[/.\-](?:19|20)\d{2}\b`, null, "DD/MM/YYYY", DOB_CONTEXT, true),
         p(EntityType.DATE_OF_DEATH, String.raw `\b(?:0[1-9]|[12][0-9]|3[01])[/.\-](?:0[1-9]|1[0-2])[/.\-](?:19|20)\d{2}\b`, null, "", DATE_OF_DEATH_CONTEXT, true),
-        p(EntityType.DOB, String.raw `\b(?:19|20)\d{2}[/.\-](?:0[1-9]|1[0-2])[/.\-](?:0[1-9]|[12][0-9]|3[01])\b`, null, "ISO YYYY-MM-DD", DOB_ISO_CONTEXT, true),
+        p(EntityType.DOB, String.raw `\b(?:19|20)\d{2}[/.\-](?:0[1-9]|1[0-2])[/.\-](?:0[1-9]|[12][0-9]|3[01])\b`, null, "ISO YYYY-MM-DD", DOB_CONTEXT, true),
     ],
 };
 const NL = {
@@ -248,7 +285,7 @@ const NL = {
         p(EntityType.NATIONAL_ID, String.raw `\b[0-9]{9}\b`, "bsn"),
         p(EntityType.NATIONAL_ID, String.raw `\b[0-9]{2,4}\.[0-9]{2,4}\.[0-9]{2,4}\b`, "bsn"),
         p(EntityType.IBAN, String.raw `\bNL\d{2}\s?[A-Z]{4}\s?\d{4}\s?\d{4}\s?\d{2}\b`, "iban"),
-        p(EntityType.VAT, String.raw `\bNL\d{9}B\d{2}\b`, "vat_nl"),
+        p(EntityType.VAT, String.raw `\bNL[\s.]?\d{9}B\d{2}\b`, "vat_nl"),
         p(EntityType.CHAMBER_OF_COMMERCE, String.raw `\b\d{8}\b`, "kvk", "", ["KvK", "Kamer van Koophandel", "kvk-nummer", "KVK-nummer", "handelsregister", "chamber of commerce"], true),
         p(EntityType.PHONE, String.raw `\b0[1-9]\d[\s\-]?\d{3}[\s\-]?\d{4}\b`),
         p(EntityType.PHONE, String.raw `\b0[1-9]\d{2}[\s\-]?\d{6,7}\b`),
@@ -275,7 +312,7 @@ const BE = {
         p(EntityType.PASSPORT, String.raw `\b[A-Z]{2}\d{6}\b`, null, "", PASSPORT_CONTEXT, true),
         p(EntityType.DRIVERS_LICENSE, String.raw `\b\d{10}\b`, null, "", ["rijbewijs", "permis de conduire", "driving licence", "driving license", "rijbewijsnummer"], true),
         p(EntityType.LICENSE_PLATE, String.raw `\b[12][\-\s]?[A-Z]{3}[\-\s]?\d{3}\b`),
-        p(EntityType.POSTAL_CODE, String.raw `\b[1-9]\d{3}\b`, null, "", ["postcode", "code postal", "postnummer", "postal code", "zip", "B-", "adres", "adresse", "wonende", "woonplaats", "rue", "straat", "laan", "avenue", "boulevard", "plein", "steenweg", "chaussée", "domicilié", "gedomicilieerd", "Levering:", "siège"], true),
+        p(EntityType.POSTAL_CODE, String.raw `\b[1-9]\d{3}\b`, null, "", ["postcode", "code postal", "postnummer", "postal code", "zip", "B-", "adres", "adresse", "wonende", "woonplaats", "domicile", "rue", "straat", "laan", "avenue", "boulevard", "plein", "steenweg", "chaussée", "domicilié", "gedomicilieerd", "Levering:", "siège"], true),
         p(EntityType.POSTAL_CODE, String.raw `(?<=, )[1-9]\d{3}(?= [A-Z])`),
         p(EntityType.HEALTHCARE_PROVIDER, String.raw `\b\d{1}[\-.]?\d{5}[\-.]?\d{2}[\-.]?\d{3}\b`, null, "", ["RIZIV", "INAMI", "arts", "médecin", "zorgverlener", "prestataire", "dokter", "doctor"], true),
     ],
@@ -300,7 +337,7 @@ const DE = {
         // separates the city code: "B-AB 1234", "M-XY 99".
         p(EntityType.LICENSE_PLATE, String.raw `\b[A-ZÄÖÜ]{1,3}[\s\-][A-Z]{1,2}[\s\-]?\d{1,4}[EH]?\b`),
         p(EntityType.POSTAL_CODE, String.raw `\b\d{5}\b`, null, "", ["PLZ", "Postleitzahl", "postal code", "postcode", "Anschrift", "Adresse", "Wohnort", "Ort", "Stadt", "wohnt", "wohnhaft", "Straße", "Str.", "Weg", "Platz", "Allee", "Ring"], true),
-        p(EntityType.CHAMBER_OF_COMMERCE, String.raw `\bHR[AB]\s?\d{4,6}\b`),
+        p(EntityType.CHAMBER_OF_COMMERCE, String.raw `\bHR[AB]\s?\d{4,6}(?:\s[A-Z]\b)?`),
         p(EntityType.SSN, String.raw `\b\d{2}[\s]?\d{6}[\s]?[A-Z][\s]?\d{3}\b`, null, "", ["Rentenversicherung", "RV-Nummer", "Sozialversicherung", "SV-Nummer", "Versicherungsnummer", "Rentenversicherungsnummer", "Sozialversicherungsnummer", "SVNR", "SV-Nr", "RVNR"], true),
         p(EntityType.TAX_ID, String.raw `\b\d{2,3}/\d{3,4}/\d{4,5}\b`, null, "", ["Steuernummer", "St.-Nr", "StNr", "Finanzamt", "Steuer"], true),
         p(EntityType.TAX_ID, String.raw `\b\d{13}\b`, null, "", ["Steuernummer", "St.-Nr", "StNr", "Finanzamt", "Steuer"], true),
@@ -313,7 +350,7 @@ const AT = {
     patterns: [
         p(EntityType.NATIONAL_ID, String.raw `\b\d{4}\s?\d{6}\b`, "austrian_svnr"),
         p(EntityType.IBAN, String.raw `\bAT\d{2}\s?\d{4}\s?\d{4}\s?\d{4}\s?\d{4}\b`, "iban"),
-        p(EntityType.VAT, String.raw `\bATU\d{8}\b`),
+        p(EntityType.VAT, String.raw `\bATU[\s.]?\d{8}\b`),
         p(EntityType.PHONE, String.raw `\b0\d{3,4}[\s\-]?\d{5,8}\b`),
         // Short area codes (Vienna is "01") split across three groups:
         // "01 53460 2215". The pattern above requires 3-4 digits after the trunk
@@ -354,7 +391,7 @@ const FR = {
         p(EntityType.PHONE, String.raw `\+33\s?[1-9][\s.\-]?\d{2}[\s.\-]?\d{2}[\s.\-]?\d{2}[\s.\-]?\d{2}`),
         p(EntityType.PASSPORT, String.raw `\b\d{2}[A-Z]{2}\d{5}\b`, null, "", PASSPORT_CONTEXT, true),
         p(EntityType.LICENSE_PLATE, String.raw `\b[A-Z]{2}[\-\s]?\d{3}[\-\s]?[A-Z]{2}\b`),
-        p(EntityType.POSTAL_CODE, String.raw `\b(?:0[1-9]|[1-8]\d|9[0-5]|97[1-6])\d{3}\b`, null, "", ["code postal", "CP", "postal code", "postcode", "adresse", "domicilié", "résidant", "rue", "avenue", "boulevard", "place", "chemin", "allée", "impasse", "ville"], true),
+        p(EntityType.POSTAL_CODE, String.raw `\b(?:0[1-9]|[1-8]\d|9[0-5]|97[1-6])\d{3}\b`, null, "", ["code postal", "CP", "postal code", "postcode", "adresse", "domicile", "domicilié", "résidant", "rue", "avenue", "boulevard", "place", "chemin", "allée", "impasse", "ville"], true),
         p(EntityType.CHAMBER_OF_COMMERCE, String.raw `\b\d{3}\s?\d{3}\s?\d{3}\b`, null, "", ["SIREN", "siren", "RCS", "entreprise", "immatricul", "numéro d'entreprise", "registre du commerce"], true),
         p(EntityType.CHAMBER_OF_COMMERCE, String.raw `\b\d{3}\s?\d{3}\s?\d{3}\s?\d{5}\b`, null, "", ["SIRET", "siret", "établissement", "immatricul"], true),
         p(EntityType.TAX_ID, String.raw `\b\d{13}\b`, null, "", ["numéro fiscal", "SPI", "référence fiscale", "avis d'impôt", "impôt sur le revenu", "déclaration fiscale", "fiscal"], true),
@@ -387,9 +424,9 @@ const SE = {
         p(EntityType.PHONE, String.raw `\b0\d{1,3}[\-\s]?\d{2,3}[\s]?\d{2,3}[\s]?\d{2}\b`),
         p(EntityType.PHONE, String.raw `\+46\s?\d{1,3}[\s\-]?\d{2,3}[\s\-]?\d{2,3}[\s\-]?\d{2}`),
         p(EntityType.LICENSE_PLATE, String.raw `\b[A-Z]{3}\s?\d{2}[A-Z0-9]\b`),
-        p(EntityType.POSTAL_CODE, String.raw `\b\d{3}\s?\d{2}\b`, null, "", ["postnummer", "postort", "postkod", "adress", "bostadsadress", "gatuadress", "boende", "gatan", "vägen", "gata", "väg", "allé", "plats", "torg", "stigen", "Bostadsadress"], true),
-        p(EntityType.POSTAL_CODE, String.raw `\b\d{4}\s\d{2}\b`, null, "", ["postnummer", "postort", "postkod", "adress", "bostadsadress", "gatuadress", "boende", "gatan", "vägen", "gata", "väg", "allé"], true),
-        p(EntityType.POSTAL_CODE, String.raw `\b\d{6}\b`, null, "", ["postnummer", "postort", "postkod", "adress", "bostadsadress", "gatuadress", "boende", "gatan", "vägen", "gata", "väg", "allé"], true),
+        p(EntityType.POSTAL_CODE, String.raw `\b\d{3}\s?\d{2}\b`, null, "", ["postnummer", "postort", "postkod", "adress", "Postal:", "Address:", "bostadsadress", "gatuadress", "boende", "gatan", "vägen", "gata", "väg", "allé", "plats", "torg", "stigen", "Bostadsadress"], true),
+        p(EntityType.POSTAL_CODE, String.raw `\b\d{4}\s\d{2}\b`, null, "", ["postnummer", "postort", "postkod", "adress", "Postal:", "Address:", "bostadsadress", "gatuadress", "boende", "gatan", "vägen", "gata", "väg", "allé"], true),
+        p(EntityType.POSTAL_CODE, String.raw `\b\d{6}\b`, null, "", ["postnummer", "postort", "postkod", "adress", "Postal:", "Address:", "bostadsadress", "gatuadress", "boende", "gatan", "vägen", "gata", "väg", "allé"], true),
         p(EntityType.CHAMBER_OF_COMMERCE, String.raw `\b\d{6}-?\d{4}\b`, null, "", ["organisationsnummer", "org.nr", "org nr", "Bolagsverket", "registreringsnummer"], true),
     ],
 };
@@ -405,7 +442,7 @@ const NO = {
         p(EntityType.PHONE, String.raw `\b[2-9]\d\s?\d{2}\s?\d{2}\s?\d{2}\b`),
         p(EntityType.PHONE, String.raw `\+47\s?\d{2}\s?\d{2}\s?\d{2}\s?\d{2}`),
         p(EntityType.LICENSE_PLATE, String.raw `\b[A-Z]{2}\s?\d{5}\b`),
-        p(EntityType.POSTAL_CODE, String.raw `\b\d{4}\b`, null, "", ["postnummer", "poststed", "postnr", "adresse", "bostedsadresse", "gate", "vei", "veien", "gata", "plass", "stien", "allé"], true),
+        p(EntityType.POSTAL_CODE, String.raw `\b\d{4}\b`, null, "", ["postnummer", "poststed", "postnr", "adresse", "bostedsadresse", "gate", "vei", "veien", "gata", "plass", "stien", "allé", "Postal:", "Address:"], true),
     ],
 };
 const FI = {
@@ -414,7 +451,7 @@ const FI = {
     patterns: [
         p(EntityType.NATIONAL_ID, String.raw `\b\d{6}[-+ABCDEFYXWVU]\d{3}[0-9A-FHJK-NPR-Y]\b`, "finnish_hetu"),
         p(EntityType.IBAN, String.raw `\bFI\d{2}\s?\d{4}\s?\d{4}\s?\d{4}\s?\d{2}\b`, "iban"),
-        p(EntityType.VAT, String.raw `\bFI\d{8}\b`),
+        p(EntityType.VAT, String.raw `\bFI[\s.]?\d{8}\b`),
         p(EntityType.CHAMBER_OF_COMMERCE, String.raw `\b\d{7}-\d\b`, "finnish_business_id"),
         p(EntityType.PHONE, String.raw `\b0\d{1,2}\s?\d{3,4}\s?\d{3,4}\b`),
         p(EntityType.PHONE, String.raw `\+358\s?\d{1,3}[\s\-]?\d{3,4}[\s\-]?\d{3,4}`),
@@ -432,7 +469,7 @@ const IS = {
         p(EntityType.PHONE, String.raw `\b[3-9]\d{6}\b`, null, "", ["sími", "símanúmer", "farsími", "gsm", "phone", "tel", "hringja", "hringdu", "ná í", "nás á", "nás", "SMS", "sent to"], true),
         p(EntityType.PHONE, String.raw `\+354\s?\d{3}\s?\d{4}`),
         p(EntityType.LICENSE_PLATE, String.raw `\b[A-Z]{2}\s?\d{3}\b`, null, "", ["skráningarnúmer", "bílnúmer", "ökutæki", "bifreið", "plata"], true),
-        p(EntityType.POSTAL_CODE, String.raw `\b[1-9]\d{2}\b`, null, "", ["póstnúmer", "póstfang", "staður", "heimilisfang"], true),
+        p(EntityType.POSTAL_CODE, String.raw `\b[1-9]\d{2}\b`, null, "", ["póstnúmer", "póstfang", "staður", "heimilisfang", "Postal:", "Address:"], true),
     ],
 };
 const IT = {
@@ -442,7 +479,7 @@ const IT = {
         p(EntityType.NATIONAL_ID, String.raw `\b[A-Z]{6}\d{2}[ABCDEHLMPRST]\d{2}[A-Z]\d{3}[A-Z]\b`, "italian_cf"),
         p(EntityType.IBAN, String.raw `\bIT\d{2}\s?[A-Z]\d{3}\s?\d{4}\s?\d{4}\s?\d{4}\s?\d{4}\s?\d{3}\b`, "iban"),
         p(EntityType.IBAN, String.raw `\bIT\d{2}[A-Z]\d{22}\b`, "iban"),
-        p(EntityType.VAT, String.raw `\bIT\d{11}\b`),
+        p(EntityType.VAT, String.raw `\bIT[\s.]?\d{11}\b`),
         p(EntityType.PHONE, String.raw `\b3\d{2}[\s\-]?\d{3}[\s\-]?\d{4}\b`),
         p(EntityType.PHONE, String.raw `\b0\d{1,3}[\s\-]?\d{6,8}\b`),
         p(EntityType.PHONE, String.raw `\b0\d{1,2}[\s\-]\d{3}[\s\-]\d{4}\b`),
@@ -458,7 +495,7 @@ const ES = {
         p(EntityType.NATIONAL_ID, String.raw `\b\d{8}[A-Z]\b`, "spanish_dni"),
         p(EntityType.NATIONAL_ID, String.raw `\b[XYZ]\d{7}[A-Z]\b`, "spanish_nie"),
         p(EntityType.IBAN, String.raw `\bES\d{2}\s?\d{4}\s?\d{4}\s?\d{4}\s?\d{4}\s?\d{4}\b`, "iban"),
-        p(EntityType.VAT, String.raw `\bES[A-Z0-9]\d{7}[A-Z0-9]\b`),
+        p(EntityType.VAT, String.raw `\bES[\s.]?[A-Z0-9]\d{7}[A-Z0-9]\b`),
         p(EntityType.PHONE, String.raw `\b[679]\d{2}[\s\-]?\d{3}[\s\-]?\d{3}\b`),
         p(EntityType.PHONE, String.raw `\b[679]\d{8}\b`),
         p(EntityType.PHONE, String.raw `\b[679]\d[\s\-]?\d{3}[\s\-]?\d{2}[\s\-]?\d{2}\b`),
@@ -481,7 +518,7 @@ const PT = {
         p(EntityType.NATIONAL_ID, String.raw `\b[1-35-9]\d{8}\b`, "portuguese_nif"),
         p(EntityType.NATIONAL_ID, String.raw `\b[1-35-9]\d{2}[\s.]\d{3}[\s.]\d{3}\b`, "portuguese_nif"),
         p(EntityType.IBAN, String.raw `\bPT\d{2}\s?\d{4}\s?\d{4}\s?\d{4}\s?\d{4}\s?\d{5}\b`, "iban"),
-        p(EntityType.VAT, String.raw `\bPT\d{9}\b`),
+        p(EntityType.VAT, String.raw `\bPT[\s.]?\d{9}\b`),
         p(EntityType.PHONE, String.raw `\b[29]\d{2}[\s\-]?\d{3}[\s\-]?\d{3}\b`),
         p(EntityType.PHONE, String.raw `\b[29]\d{8}\b`),
         p(EntityType.PHONE, String.raw `\+351\s?[29]\d{2}[\s\-]?\d{3}[\s\-]?\d{3}`),
@@ -514,7 +551,7 @@ const PL = {
         p(EntityType.TAX_ID, String.raw `\b\d{3}-?\d{3}-?\d{2}-?\d{2}\b`, "polish_nip"),
         p(EntityType.IBAN, String.raw `\bPL\d{2}\s?\d{4}\s?\d{4}\s?\d{4}\s?\d{4}\s?\d{4}\s?\d{4}\b`, "iban"),
         p(EntityType.IBAN, String.raw `\bPL\d{26}\b`, "iban"),
-        p(EntityType.VAT, String.raw `\bPL\d{10}\b`),
+        p(EntityType.VAT, String.raw `\bPL[\s.]?\d{10}\b`),
         p(EntityType.PHONE, String.raw `\b[5-8]\d{2}[\s\-]?\d{3}[\s\-]?\d{3}\b`),
         p(EntityType.PHONE, String.raw `\b[5-8]\d{8}\b`),
         p(EntityType.PHONE, String.raw `\+48\s?[5-8]\d{2}[\s\-]?\d{3}[\s\-]?\d{3}`),
@@ -528,7 +565,7 @@ const IE = {
         p(EntityType.NATIONAL_ID, String.raw `\b\d{7}[A-W][ABWTXZ]?\b`, "irish_pps"),
         p(EntityType.IBAN, String.raw `\bIE\d{2}\s?[A-Z]{4}\s?\d{4}\s?\d{4}\s?\d{4}\s?\d{2}\b`, "iban"),
         p(EntityType.IBAN, String.raw `\bIE\d{2}[A-Z]{4}\d{14}\b`, "iban"),
-        p(EntityType.VAT, String.raw `\bIE\d{7}[A-Z]\b`),
+        p(EntityType.VAT, String.raw `\bIE[\s.]?\d{7}[A-Z]\b`),
         p(EntityType.PHONE, String.raw `\b0[89]\d[\s\-]?\d{3}[\s\-]?\d{4}\b`),
         p(EntityType.PHONE, String.raw `\b0[89]\d{8}\b`),
         p(EntityType.PHONE, String.raw `\b0[1-9]\d{0,2}[\s\-]?\d{5,8}\b`),
@@ -544,7 +581,7 @@ const UK = {
         p(EntityType.HEALTH_INSURANCE, String.raw `\b\d{3}\s?\d{3}\s?\d{4}\b`, "uk_nhs", "", ["NHS", "NHS number", "health number", "NHS no"], true),
         p(EntityType.IBAN, String.raw `\bGB\d{2}\s?[A-Z]{4}\s?\d{4}\s?\d{4}\s?\d{4}\s?\d{2}\b`, "iban"),
         p(EntityType.IBAN, String.raw `\bGB\d{2}[A-Z]{4}\d{14}\b`, "iban"),
-        p(EntityType.VAT, String.raw `\bGB\d{9}\b`),
+        p(EntityType.VAT, String.raw `\bGB[\s.]?\d{3}[\s.]?\d{4}[\s.]?\d{2}\b`),
         p(EntityType.PHONE, String.raw `\b07[4-9]\d{2}[\s\-]?\d{3}[\s\-]?\d{3}\b`),
         p(EntityType.PHONE, String.raw `\b07[4-9]\d{8}\b`),
         p(EntityType.PHONE, String.raw `\b0[12]\d{2}[\s\-]?\d{3}[\s\-]?\d{4}\b`),
@@ -560,7 +597,7 @@ const BG = {
     patterns: [
         p(EntityType.NATIONAL_ID, String.raw `\b\d{10}\b`, "bulgarian_egn"),
         p(EntityType.IBAN, String.raw `\bBG\d{2}\s?[A-Z]{4}\s?\d{4}\s?\d{4}\s?\d{4}\s?\d{2}\b`, "iban"),
-        p(EntityType.VAT, String.raw `\bBG\d{9,10}\b`),
+        p(EntityType.VAT, String.raw `\bBG[\s.]?\d{9,10}\b`),
         p(EntityType.PHONE, String.raw `\b0?[89][789][\s\-]?\d{3}[\s\-]?\d{3,4}\b`),
         p(EntityType.PHONE, String.raw `\b0?[89][789]\d{7,8}\b`),
         p(EntityType.PHONE, String.raw `\+359\s?[89][789][\s\-]?\d{3}[\s\-]?\d{3,4}`),
@@ -575,7 +612,7 @@ const CY = {
         p(EntityType.NATIONAL_ID, String.raw `\b\d{8}[A-Z]\b`),
         p(EntityType.IBAN, String.raw `\bCY\d{2}\s?\d{4}\s?\d{4}\s?\d{4}\s?\d{4}\s?\d{4}\s?\d{4}\b`, "iban"),
         p(EntityType.IBAN, String.raw `\bCY\d{26}\b`, "iban"),
-        p(EntityType.VAT, String.raw `\bCY\d{8}[A-Z]\b`),
+        p(EntityType.VAT, String.raw `\bCY[\s.]?\d{8}[A-Z]\b`),
         p(EntityType.PHONE, String.raw `\b[29]\d{3}[\s\-]?\d{4}\b`),
         p(EntityType.PHONE, String.raw `\b[29]\d[\s\-]?\d{3}[\s\-]?\d{3}\b`),
         p(EntityType.PHONE, String.raw `\b[29]\d{7}\b`),
@@ -591,7 +628,7 @@ const CZ = {
         p(EntityType.NATIONAL_ID, String.raw `\b\d{6}/?\d{3,4}\b`, "czech_birth_number"),
         p(EntityType.IBAN, String.raw `\bCZ\d{2}\s?\d{4}\s?\d{4}\s?\d{4}\s?\d{4}\s?\d{4}\b`, "iban"),
         p(EntityType.IBAN, String.raw `\bCZ\d{22}\b`, "iban"),
-        p(EntityType.VAT, String.raw `\bCZ\d{8,10}\b`),
+        p(EntityType.VAT, String.raw `\bCZ[\s.]?\d{8,10}\b`),
         p(EntityType.PHONE, String.raw `\b[67]\d{2}[\s\-]?\d{3}[\s\-]?\d{3}\b`),
         p(EntityType.PHONE, String.raw `\b[67]\d{8}\b`),
         p(EntityType.PHONE, String.raw `\+420\s?[67]\d{2}[\s\-]?\d{3}[\s\-]?\d{3}`),
@@ -605,7 +642,7 @@ const SK = {
         p(EntityType.NATIONAL_ID, String.raw `\b\d{6}/?\d{3,4}\b`, "czech_birth_number"),
         p(EntityType.IBAN, String.raw `\bSK\d{2}\s?\d{4}\s?\d{4}\s?\d{4}\s?\d{4}\s?\d{4}\b`, "iban"),
         p(EntityType.IBAN, String.raw `\bSK\d{22}\b`, "iban"),
-        p(EntityType.VAT, String.raw `\bSK\d{10}\b`),
+        p(EntityType.VAT, String.raw `\bSK[\s.]?\d{10}\b`),
         p(EntityType.PHONE, String.raw `\b0?9\d{2}[\s\-]?\d{3}[\s\-]?\d{3}\b`),
         p(EntityType.PHONE, String.raw `\+421\s?9\d{2}[\s\-]?\d{3}[\s\-]?\d{3}`),
         p(EntityType.POSTAL_CODE, String.raw `\b\d{3}\s?\d{2}\b`, null, "", ["PSČ", "poštové", "adresa", "ulica", "Postal:", "Address:"], true),
@@ -618,7 +655,7 @@ const EE = {
         p(EntityType.NATIONAL_ID, String.raw `\b[1-6]\d{10}\b`, "estonian_id"),
         p(EntityType.IBAN, String.raw `\bEE\d{2}\s?\d{4}\s?\d{4}\s?\d{4}\s?\d{4}\b`, "iban"),
         p(EntityType.IBAN, String.raw `\bEE\d{18}\b`, "iban"),
-        p(EntityType.VAT, String.raw `\bEE\d{9}\b`),
+        p(EntityType.VAT, String.raw `\bEE[\s.]?\d{9}\b`),
         p(EntityType.PHONE, String.raw `\b5\d{3}[\s\-]?\d{4}\b`),
         p(EntityType.PHONE, String.raw `\b5\d{7}\b`),
         p(EntityType.PHONE, String.raw `\+372\s?5\d{3}[\s\-]?\d{3,4}`),
@@ -632,7 +669,7 @@ const LT = {
         p(EntityType.NATIONAL_ID, String.raw `\b[1-6]\d{10}\b`, "estonian_id"),
         p(EntityType.IBAN, String.raw `\bLT\d{2}\s?\d{4}\s?\d{4}\s?\d{4}\s?\d{4}\b`, "iban"),
         p(EntityType.IBAN, String.raw `\bLT\d{18}\b`, "iban"),
-        p(EntityType.VAT, String.raw `\bLT\d{9,12}\b`),
+        p(EntityType.VAT, String.raw `\bLT[\s.]?\d{9,12}\b`),
         p(EntityType.PHONE, String.raw `\b6\d{2}[\s\-]?\d{2}[\s\-]?\d{3}\b`),
         p(EntityType.PHONE, String.raw `\b6\d{7}\b`),
         p(EntityType.PHONE, String.raw `\b8[\s\-]?6\d{2}[\s\-]?\d{2}[\s\-]?\d{3}\b`),
@@ -648,7 +685,7 @@ const LV = {
         p(EntityType.NATIONAL_ID, String.raw `\b(?:0[1-9]|[12]\d|3[01])(?:0[1-9]|1[0-2])\d{7}\b`, null, "", ["personas kods", "PK", "identifikācijas"], true),
         p(EntityType.IBAN, String.raw `\bLV\d{2}\s?[A-Z]{4}\s?\d{4}\s?\d{4}\s?\d{4}\s?\d\b`, "iban"),
         p(EntityType.IBAN, String.raw `\bLV\d{2}[A-Z]{4}\d{13}\b`, "iban"),
-        p(EntityType.VAT, String.raw `\bLV\d{11}\b`),
+        p(EntityType.VAT, String.raw `\bLV[\s.]?\d{11}\b`),
         p(EntityType.PHONE, String.raw `\b[2]\d{3}[\s\-]?\d{4}\b`),
         p(EntityType.PHONE, String.raw `\b[2]\d{7}\b`),
         p(EntityType.PHONE, String.raw `\+371\s?[2]\d{3}[\s\-]?\d{3,4}`),
@@ -663,7 +700,7 @@ const HR = {
         p(EntityType.IBAN, String.raw `\bHR\d{2}\s?\d{4}\s?\d{4}\s?\d{4}\s?\d{4}\s?\d{3}\b`, "iban"),
         p(EntityType.IBAN, String.raw `\bHR\d{2}\s\d{4}\s\d{4}\s\d{4}\s\d{5}\b`, "iban"),
         p(EntityType.IBAN, String.raw `\bHR\d{19}\b`, "iban"),
-        p(EntityType.VAT, String.raw `\bHR\d{11}\b`),
+        p(EntityType.VAT, String.raw `\bHR[\s.]?\d{11}\b`),
         p(EntityType.PHONE, String.raw `\b0?9[12579][\s\-]?\d{3}[\s\-]?\d{3,4}\b`),
         p(EntityType.PHONE, String.raw `\b0?9[12579]\d{7,8}\b`),
         p(EntityType.PHONE, String.raw `\+385\s?9[12579][\s\-]?\d{3}[\s\-]?\d{3,4}`),
@@ -677,7 +714,7 @@ const HU = {
         p(EntityType.NATIONAL_ID, String.raw `\b\d{9}\b`, "hungarian_taj"),
         p(EntityType.NATIONAL_ID, String.raw `\b\d{3}\s\d{3}\s\d{3}\b`, "hungarian_taj"),
         p(EntityType.TAX_ID, String.raw `\b8\d{9}\b`, null, "", ["adóazonosító", "adószám", "adóazonosító jel"], true),
-        p(EntityType.VAT, String.raw `\bHU\d{8,11}\b`, null, "", ["adószám", "ÁFA"]),
+        p(EntityType.VAT, String.raw `\bHU[\s.]?\d{8,11}\b`, null, "", ["adószám", "ÁFA"]),
         p(EntityType.IBAN, String.raw `\bHU\d{2}\s?\d{4}\s?\d{4}\s?\d{4}\s?\d{4}\s?\d{4}\s?\d{4}\b`, "iban"),
         p(EntityType.IBAN, String.raw `\bHU\d{26}\b`, "iban"),
         p(EntityType.PHONE, String.raw `\b06\s?\d{1,2}[\s\-]?\d{3}[\s\-]?\d{3,4}\b`),
@@ -692,7 +729,7 @@ const RO = {
     patterns: [
         p(EntityType.NATIONAL_ID, String.raw `\b[1-8]\d{12}\b`, "romanian_cnp"),
         p(EntityType.IBAN, String.raw `\bRO\d{2}[A-Z]{4}\d{16}\b`, "iban"),
-        p(EntityType.VAT, String.raw `\bRO\d{2,10}\b`),
+        p(EntityType.VAT, String.raw `\bRO[\s.]?\d{2,10}\b`),
         p(EntityType.PHONE, String.raw `\b0?7[2-9]\d{1}[\s\-]?\d{3}[\s\-]?\d{3,4}\b`),
         p(EntityType.PHONE, String.raw `\b0?7[2-9]\d{7,8}\b`),
         p(EntityType.PHONE, String.raw `\+40\s?7[2-9]\d{1}[\s\-]?\d{3}[\s\-]?\d{3,4}`),
@@ -707,7 +744,7 @@ const EL = {
         p(EntityType.NATIONAL_ID, String.raw `\b\d{11}\b`, "greek_amka"),
         p(EntityType.IBAN, String.raw `\bGR\d{2}\s?\d{4}\s?\d{4}\s?\d{4}\s?\d{4}\s?\d{4}\s?\d{3}\b`, "iban"),
         p(EntityType.IBAN, String.raw `\bGR\d{25}\b`, "iban"),
-        p(EntityType.VAT, String.raw `\bEL\d{9}\b`),
+        p(EntityType.VAT, String.raw `\bEL[\s.]?\d{9}\b`),
         p(EntityType.PHONE, String.raw `\b69\d[\s\-]?\d{3,4}[\s\-]?\d{4}\b`),
         p(EntityType.PHONE, String.raw `\b69\d{8}\b`),
         p(EntityType.PHONE, String.raw `\b2\d{1,3}[\s\-]?\d{3,4}[\s\-]?\d{3,4}\b`),
@@ -722,7 +759,7 @@ const SI = {
         p(EntityType.NATIONAL_ID, String.raw `\b\d{13}\b`, "slovenian_emso"),
         p(EntityType.IBAN, String.raw `\bSI\d{2}\s?\d{4}\s?\d{4}\s?\d{4}\s?\d{3}\b`, "iban"),
         p(EntityType.IBAN, String.raw `\bSI\d{17}\b`, "iban"),
-        p(EntityType.VAT, String.raw `\bSI\d{8}\b`),
+        p(EntityType.VAT, String.raw `\bSI[\s.]?\d{8}\b`),
         p(EntityType.PHONE, String.raw `\b0?[34567]\d{1}[\s\-]?\d{3}[\s\-]?\d{3}\b`),
         p(EntityType.PHONE, String.raw `\b0?[34567]\d{7}\b`),
         p(EntityType.PHONE, String.raw `\+386\s?[34567]\d{1}[\s\-]?\d{3}[\s\-]?\d{3}`),
@@ -737,7 +774,7 @@ const MT = {
         p(EntityType.NATIONAL_ID, String.raw `\b\d{5,7}[MGAPLHBZ]\b`),
         p(EntityType.IBAN, String.raw `\bMT\d{2}\s?[A-Z]{4}\s?\d{4}\s?\d{4}\s?[A-Z0-9]{4}\s?[A-Z0-9]{4}\s?[A-Z0-9]{4}\s?[A-Z0-9]{3}\b`, "iban"),
         p(EntityType.IBAN, String.raw `\bMT\d{2}[A-Z]{4}\d{5}[A-Z0-9]{18}\b`, "iban"),
-        p(EntityType.VAT, String.raw `\bMT\d{8}\b`),
+        p(EntityType.VAT, String.raw `\bMT[\s.]?\d{8}\b`),
         p(EntityType.PHONE, String.raw `\b[79]\d{3}[\s\-]?\d{4}\b`),
         p(EntityType.PHONE, String.raw `\b[79]\d{7}\b`),
         p(EntityType.PHONE, String.raw `\+356\s?[79]\d{3}[\s\-]?\d{3,4}`),