@whiteslove/parsing-lexicon 0.10.5 → 0.10.7

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@whiteslove/parsing-lexicon",
-  "version": "0.10.5",
+  "version": "0.10.7",
   "description": "Shared deterministic multilingual parsing lexicon for WhitesLove housing and hiring services",
   "repository": {
     "type": "git",

package/src/housing-address.js

@@ -28,6 +28,8 @@ const POSTFIX_STREET_MARKER = `(?:${combinedMarkerPattern(STREET_POSTFIX_MARKERS
 const POSTFIX_STREET_TYPE = `(?:${combinedMarkerPattern(STREET_TYPE_MARKERS)})`;
 const HOUSE_MARKER = `(?:${combinedMarkerPattern(HOUSE_MARKERS)})`;
 const BUILDING_MARKER = `(?:${combinedMarkerPattern(BUILDING_MARKERS)})`;
+// Uzbek/Kazakh put the house word after the number: "12-uy", "45 uy".
+const HOUSE_NUMBER_SUFFIX = String.raw`(?:\s*-?\s*(?:uy|уй|үй)(?![\p{L}\p{N}]))?`;
 const NUMBER_TOKEN = String.raw`\d{1,5}(?:[-\/]?[\p{L}]\d{0,4})?(?:[\/-]\d{1,4}(?:[-\/]?[\p{L}]\d{0,4})?){0,2}`;
 const STREET_WORD = String.raw`[\p{L}'’.-]{2,48}`;
 // Common post-Soviet street names lead with a bare numeral ("8 Марта",
@@ -338,12 +340,31 @@ function attachCatalogReferences(parsed, options) {
   return Object.keys(geoEntities).length ? Object.freeze({ ...parsed, geoEntities }) : parsed;
 }
 
+// "Sergeli 5 104": numbered massif followed by the house number.
+function tashkentAreaThenHouseAddress(text) {
+  const re = /(?:^|[^\p{L}\p{N}_])([\p{L}'’ʻʼ-]+)\s+(\d{1,2})\s*[,;]?\s+(?:(?:dom|дом|uy|уй)\s*)?(\d{1,4})([\p{L}])?(?=$|[^\p{L}\p{N}_/])/giu;
+  for (const match of text.matchAll(re)) {
+    const quarterNumber = Number(match[2]);
+    const district = Object.keys(TASHKENT_NUMBERED_AREA_ALIASES).find((canonical) => (
+      matchTashkentNumberedArea(`${match[1]} ${quarterNumber}`, canonical)?.number === quarterNumber
+    ));
+    if (!district) continue;
+    return Object.freeze({
+      ...result(null, null, `${Number(match[3])}${match[4] ? match[4].toUpperCase() : ''}`, null,
+        scoreAddressConfidence(text, { source: 'structured', hasHouse: true })),
+      district,
+      quarter: Object.freeze({ number: quarterNumber, suffix: matchTashkentNumberedArea(`${match[1]} ${quarterNumber}`, district)?.suffix || '' }),
+    });
+  }
+  return null;
+}
+
 function tashkentMassifHouseAddress(value) {
   const text = String(value ?? '');
   const match = text.match(
     /(?:^|[^\p{L}\p{N}_])(\d{1,2})\s+(?:mavze(?:si)?|мавзе(?:си)?)\s+(\d{1,5})\s*([\p{L}])?\s*(?:dom|дом|uy|уй)(?=$|[^\p{L}\p{N}_])/iu,
   );
-  if (!match) return null;
+  if (!match) return tashkentAreaThenHouseAddress(text);
 
   const quarterNumber = Number(match[1]);
   const district = Object.keys(TASHKENT_NUMBERED_AREA_ALIASES).find((canonical) => {
@@ -371,7 +392,7 @@ function tashkentMassifHouseAddress(value) {
 }
 
 function splitAddressTail(raw) {
-  const text = clean(raw);
+  const text = clean(raw).replace(new RegExp(`(${NUMBER_TOKEN})${HOUSE_NUMBER_SUFFIX}(?=\\s*[,;]|\\s*$)`, 'iu'), '$1');
   if (!text) return null;
 
   const buildingRe = new RegExp(`(?:\\s*[,;]?\\s*${BUILDING_MARKER}\\s*(${NUMBER_TOKEN}))\\s*$`, 'iu');
@@ -400,7 +421,7 @@ function splitAddressTail(raw) {
 function postfixTypedStreetAddress(line) {
   const suffix = line.match(new RegExp(
     `(?:^|[^\\p{L}\\p{N}])((?:${LEADING_STREET_NUMERAL}\\s+)?(?:${STREET_WORD}\\s+){0,4}${STREET_WORD}\\s+${POSTFIX_STREET_TYPE})` +
-      `\\s*[,;]?\\s*(${NUMBER_TOKEN})` +
+      `\\s*[,;]?\\s*(${NUMBER_TOKEN})${HOUSE_NUMBER_SUFFIX}` +
       `(?:\\s*[,;]?\\s*${BUILDING_MARKER}\\s*(${NUMBER_TOKEN}))?` +
       `(?=$|[^\\p{L}\\p{N}])`,
     'iu',
@@ -424,7 +445,7 @@ function prefixTypedStreetAddress(line) {
   const prefix = line.match(new RegExp(
     `(?:^|[\\s,;])${PREFIX_STREET_MARKER}\\s+` +
       `((?:${LEADING_STREET_NUMERAL}\\s+)?(?:${STREET_WORD}\\s+){0,4}${STREET_WORD})` +
-      `\\s*[,;]?\\s*(?:${HOUSE_MARKER}\\s*)?(${NUMBER_TOKEN})` +
+      `\\s*[,;]?\\s*(?:${HOUSE_MARKER}\\s*)?(${NUMBER_TOKEN})${HOUSE_NUMBER_SUFFIX}` +
       `(?:\\s*[,;]?\\s*${BUILDING_MARKER}\\s*(${NUMBER_TOKEN}))?` +
       `(?=$|[^\\p{L}\\p{N}])`,
     'iu',
@@ -529,7 +550,12 @@ function collectExplicitStreetCandidates(text) {
 
     const postfix = line.match(new RegExp(`^(.+?)\\s+(${POSTFIX_STREET_MARKER})(.*)$`, 'iu'));
     if (postfix) {
-      const tailText = clean(`${postfix[1]} ${postfix[3]}`);
+      // Whatever follows the first comma ("..., Sergeli") is a district or
+      // landmark, not part of the street/house.
+      // A building marker after the comma ("58, korpus 2") still belongs to it.
+      const afterStreet = postfix[3];
+      const cut = afterStreet.search(new RegExp(`[,;](?!\\s*${BUILDING_MARKER}\\s*${NUMBER_TOKEN})`, 'iu'));
+      const tailText = clean(`${postfix[1]} ${cut < 0 ? afterStreet : afterStreet.slice(0, cut)}`);
       const tail = splitAddressTail(tailText);
       if (tail) {
         add(result(

package/src/housing-listing-enrichment.js

@@ -209,7 +209,9 @@ export function parseHousingAudience(value) {
   if (family && students) return deepFreeze({ primary: 'family', alternatives: ['family', 'students'] });
   if (family && women) return deepFreeze({ primary: 'family', alternatives: ['family', 'women'] });
   // "Student qizlarga" narrows students to girls: the gender is the binding
-  // restriction, and studentTarget still reports the student part.
+  // primary restriction, but the student audience stays in the alternatives.
+  if (students && women && !men) return deepFreeze({ primary: 'women', alternatives: ['women', 'students'] });
+  if (students && men && !women) return deepFreeze({ primary: 'men', alternatives: ['men', 'students'] });
   if (students && !women && !men) return deepFreeze({ primary: 'students', alternatives: ['students'] });
   if (family) return deepFreeze({ primary: 'family', alternatives: ['family'] });
   if (women) return deepFreeze({ primary: 'women', alternatives: ['women'] });

package/src/housing-money.js

@@ -634,8 +634,9 @@ export function parseHousingPrice(value, context = '') {
   if (!explicit && !hasExplicitUzbekThousandScale && fallbackCurrency === 'UZS' && price != null) {
     const dailyUzbek = /(?:kunlik|sutkaga|kecha[- ]?kunduz|посуточн|суточн)/i.test(priceText);
     // A monthly rent of 100,000+ is only plausible in so'm; the USD reading
-    // turned "800 000" per person into an 800,000 USD rent.
-    const monthlyRent = dealType === 'longRent' && price >= 100_000;
+    // turned "800 000" per person into an 800,000 USD rent. When the deal type
+    // is unknown, 100,000+ is still so'm: only a known sale may be dollars.
+    const monthlyRent = dealType !== 'sale' && price >= 100_000;
     currency = price >= 1_000_000 || monthlyRent || (dailyUzbek && price >= 10_000) ? 'UZS' : 'USD';
   } else if (!explicit && currency === fallbackCurrency) {
     currency = lowLocalAmountCurrency(price, currency, dealType, priceText);

package/src/housing-structured.js

@@ -342,7 +342,8 @@ function landmarkIdentity(text, match) {
   if (numbered) {
     const localStart = before.length - numbered[0].length;
     const previous = localStart > 0 ? before[localStart - 1] : '';
-    if (!previous || !/[\p{L}\p{N}_]/u.test(previous)) {
+    // "24/7 korzinka" is an opening-hours phrase, not branch number 7.
+    if (!previous || !/[\p{L}\p{N}_/]/u.test(previous)) {
       number = Number(numbered[1]);
       rawStart = match.start - numbered[0].length;
     }