@whiteslove/parsing-lexicon 0.10.4 → 0.10.6

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@whiteslove/parsing-lexicon",
-  "version": "0.10.4",
+  "version": "0.10.6",
   "description": "Shared deterministic multilingual parsing lexicon for WhitesLove housing and hiring services",
   "repository": {
     "type": "git",

package/src/housing-address.d.ts

@@ -10,6 +10,8 @@ export type HousingAddressParts = Readonly<{
   staircase?: string;
   district?: string;
   metro?: string;
+  /** Every metro station named, primary (`metro`) first. */
+  metros?: readonly string[];
   mahalla?: string;
   geoEntities?: Readonly<Partial<Record<'street' | 'district' | 'metro' | 'mahalla' | 'residentialComplex', Readonly<{
     id: string;

package/src/housing-address.js

@@ -2,6 +2,7 @@ import {
   TASHKENT_NUMBERED_AREA_ALIASES,
   matchTashkentHousingDistrict,
   matchTashkentHousingMetro,
+  matchTashkentHousingMetros,
   matchTashkentNumberedArea,
 } from './tashkent-housing-geography.js';
 import { detectNonAddressSpans } from './semantic-spans.js';
@@ -27,6 +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",
@@ -246,10 +249,14 @@ function tashkentGeoComponents(value) {
   const text = String(value ?? '');
   const district = matchTashkentHousingDistrict(text)?.name || null;
   const metro = matchTashkentHousingMetro(text)?.name || null;
+  const metros = metro
+    ? [...new Set([metro, ...matchTashkentHousingMetros(text).map((station) => station.name)])]
+    : [];
   const mahalla = text.match(/(?:^|[^\p{L}])(\p{L}[\p{L}'’ʼ-]{1,48})\s+(?:mahalla(?:si)?|маҳалла(?:си)?|махалл[ая]|mfy)(?=$|[^\p{L}])/iu)?.[1] || null;
   return Object.freeze({
     district,
     metro,
+    metros: metros.length ? Object.freeze(metros) : null,
     mahalla: DESCRIPTIVE_MAHALLA_WORD_RE.test(mahalla || '') ? null : compactStreet(mahalla),
   });
 }
@@ -333,12 +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) => {
@@ -366,7 +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*$`, 'iu');
@@ -395,7 +421,7 @@ function splitAddressTail(raw) {
 function postfixTypedStreetAddress(line) {
   const suffix = line.match(new RegExp(
     `(?:^|[^\\p{L}\\p{N}])((?:${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}))?` +
       `(?=$|[^\\p{L}\\p{N}])`,
     'iu',
@@ -419,7 +445,7 @@ function prefixTypedStreetAddress(line) {
   const prefix = line.match(new RegExp(
     `(?:^|[\\s,;])${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',
@@ -524,7 +550,9 @@ function collectExplicitStreetCandidates(text) {
 
     const postfix = line.match(new RegExp(`^(.+?)\\s+(${POSTFIX_STREET_MARKER})(.*)$`, 'iu'));
     if (postfix) {
-      const tailText = clean(`${postfix[1]} ${postfix[3]}`);
+      // Whatever follows the first comma ("..., Sergeli") is a district or
+      // landmark, not part of the street/house.
+      const tailText = clean(`${postfix[1]} ${postfix[3].split(/[,;]/u, 1)[0]}`);
       const tail = splitAddressTail(tailText);
       if (tail) {
         add(result(

package/src/housing-listing-enrichment.d.ts

@@ -58,6 +58,8 @@ export interface HousingListingEnrichment {
   district?: string | null;
   quarter?: { number: number; suffix: string } | null;
   metro?: string | null;
+  /** Every metro station named, primary (`metro`) first. */
+  metros?: readonly string[];
   developmentArea?: string | null;
   residenceComplex?: string | null;
   address?: string | null;

package/src/housing-listing-enrichment.js

@@ -1,7 +1,7 @@
 import { deepFreeze } from './lexicon-core.js';
 import { findAllCanonical, normalizeUnicode } from './normalization.js';
 import { GENERIC_LANDMARK_TERMS } from './landmarks.js';
-import { matchTashkentHousingDistrict, matchTashkentHousingMetro, matchTashkentHousingQuarter } from './tashkent-housing-geography.js';
+import { matchTashkentHousingDistrict, matchTashkentHousingMetro, matchTashkentHousingMetros, matchTashkentHousingQuarter } from './tashkent-housing-geography.js';
 import { TASHKENT_LANDMARKS } from './tashkent-pois.js';
 import { TASHKENT_RESIDENTIAL_COMPLEXES, matchTashkentResidentialComplex } from './tashkent-residential-complexes.js';
 import { parseHousingRoomCount, parseHousingFloor, parseHousingAreas } from './housing-structured.js';
@@ -161,6 +161,16 @@ function cityMetro(text, country, city) {
   return contextualCityMatches(text, country, city, 'metro', METRO_PREFIX_RE)[0]?.canonical || null;
 }
 
+// Every station the listing names, primary first. A flat "between Nizomiy and
+// Chilonzor" must keep both rather than silently dropping the second.
+function listingMetros(text, country, city, primary) {
+  const names = [
+    ...matchTashkentHousingMetros(text).map((station) => station.name),
+    ...contextualCityMatches(text, country, city, 'metro', METRO_PREFIX_RE).map((match) => match.canonical),
+  ];
+  return Object.freeze(unique(primary ? [primary, ...names] : names));
+}
+
 function cityDevelopmentArea(text, country, city) {
   const normalizedCountry = String(country || '').toUpperCase();
   if (!['KZ', 'UZ'].includes(normalizedCountry) || !city) return null;
@@ -199,7 +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'] });
@@ -310,6 +322,7 @@ export function parseHousingListingEnrichment(value, { country = '', city = '',
   const quarter = matchTashkentHousingQuarter(text);
   const district = matchTashkentHousingDistrict(text)?.name || quarter?.district || null;
   const metro = matchTashkentHousingMetro(text)?.name || cityMetro(text, country, city) || null;
+  const metros = listingMetros(text, country, city, metro);
   const developmentArea = cityDevelopmentArea(text, country, city);
   const primaryResidentialText = withoutNearbyLocationReferences(text);
   const parsedRc = specificResidentialComplex(primaryResidentialText)
@@ -386,6 +399,7 @@ export function parseHousingListingEnrichment(value, { country = '', city = '',
     district: district || null,
     quarter: quarter ? { number: quarter.number, suffix: quarter.suffix } : null,
     metro: metro || null,
+    metros,
     developmentArea,
     residenceComplex: parsedRc || null,
     address: address.address,

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;
     }

package/src/tashkent-housing-geography.d.ts

@@ -46,4 +46,5 @@ export function hasExplicitTashkentDistrict(value: unknown, canonical: string):
 export function matchTashkentHousingDistrict(value: unknown): TashkentDistrictEntry | null;
 export function matchTashkentHousingQuarter(value: unknown): Readonly<{ district: string; number: number; suffix: string }> | null;
 export function matchTashkentHousingMetro(value: unknown): TashkentMetroEntry | null;
+export function matchTashkentHousingMetros(value: unknown): ReadonlyArray<TashkentMetroEntry>;
 export function matchTashkentHousingTransit(value: unknown): TashkentHousingLocationEntry | null;

package/src/tashkent-housing-geography.js

@@ -298,42 +298,75 @@ const METRO_NUMBERED_AREA = Object.freeze({
   Yangihayot: 'Yangihayot',
 });
 
+// Position of the first mention of `station` that really names the station,
+// or -1. An explicit marker must win even if an earlier bare occurrence shares
+// a name with a district: "Sergeli tumani, metro Sergeli" still refers to the
+// station at the second occurrence.
+function stationMentionIndex(text, station) {
+  const flags = [...new Set(`${station.re.flags.replace(/g/gu, '')}g`)].join('');
+  const matches = [...text.matchAll(new RegExp(station.re.source, flags))];
+  if (!matches.length) return -1;
+  const explicit = matches.find((match) => hasExplicitMetroContext(text, match));
+  if (explicit) return explicit.index ?? 0;
+  if (hasExplicitTashkentDistrict(text, station.name)) return -1;
+  // "Toshkent" is both a metro station and the city's own name, so a bare
+  // mention ("Toshkent shahri") is not evidence of the station the way a bare
+  // mention of any other station name would be. Require an explicit metro
+  // context for this one station specifically.
+  if (station.name === 'Toshkent') return -1;
+  if (station.name === 'Qoyliq' && QOYLIQ_MASSIF_RE.test(text)) return -1;
+  const areaCanonical = METRO_NUMBERED_AREA[station.name];
+  if (areaCanonical && matchTashkentNumberedArea(text, areaCanonical)) return -1;
+  for (const match of matches) {
+    if (hasExplicitDistrictContext(text, match) || hasExplicitAreaContext(text, match) || hasExplicitMahallaContext(text, match) || hasExplicitLandmarkContext(text, match)) continue;
+    return match.index ?? 0;
+  }
+  return -1;
+}
+
+function extraAliasMentionIndex(text, aliases) {
+  const match = text.match(aliasesToRegex(aliases));
+  if (!match) return -1;
+  if (hasExplicitDistrictContext(text, match) || hasExplicitAreaContext(text, match) || hasExplicitMahallaContext(text, match)) return -1;
+  return match.index ?? 0;
+}
+
 /** Resolve listing typos/transliterations while respecting explicit non-metro geography. */
 export function matchTashkentHousingMetro(value) {
   const text = String(value ?? '');
   if (!text) return null;
   for (const station of TASHKENT_METRO) {
-    const flags = [...new Set(`${station.re.flags.replace(/g/gu, '')}g`)].join('');
-    const matches = [...text.matchAll(new RegExp(station.re.source, flags))];
-    if (!matches.length) continue;
-    // An explicit marker must win even if an earlier bare occurrence shares a
-    // name with a district. For example, "Sergeli tumani, metro Sergeli"
-    // still refers to the station at the second occurrence.
-    if (matches.some((match) => hasExplicitMetroContext(text, match))) return station;
-    const sameNamedDistrict = hasExplicitTashkentDistrict(text, station.name);
-    for (const match of matches) {
-      if (sameNamedDistrict) continue;
-    // "Toshkent" is both a metro station and the city's own name, so a bare
-    // mention ("Toshkent shahri") is not evidence of the station the way a
-    // bare mention of any other station name would be. Require an explicit
-    // metro context for this one station specifically.
-      if (station.name === 'Toshkent') continue;
-      if (station.name === 'Qoyliq' && QOYLIQ_MASSIF_RE.test(text)) continue;
-      const areaCanonical = METRO_NUMBERED_AREA[station.name];
-      if (areaCanonical && matchTashkentNumberedArea(text, areaCanonical)) continue;
-      if (hasExplicitDistrictContext(text, match) || hasExplicitAreaContext(text, match) || hasExplicitMahallaContext(text, match) || hasExplicitLandmarkContext(text, match)) continue;
-      return station;
-    }
+    if (stationMentionIndex(text, station) !== -1) return station;
   }
   for (const [canonical, aliases] of Object.entries(EXTRA_METRO_ALIASES)) {
-    const match = text.match(aliasesToRegex(aliases));
-    if (!match) continue;
-    if (hasExplicitDistrictContext(text, match) || hasExplicitAreaContext(text, match) || hasExplicitMahallaContext(text, match)) continue;
+    if (extraAliasMentionIndex(text, aliases) === -1) continue;
     return TASHKENT_METRO.find((station) => station.name === canonical) || null;
   }
   return null;
 }
 
+/**
+ * Every Tashkent metro station a listing names, in text order. Listings often
+ * sit between stations ("Nizomiy & Chilonzor metrosiga yaqin"); the singular
+ * matcher above keeps its historical first-station contract.
+ */
+export function matchTashkentHousingMetros(value) {
+  const text = String(value ?? '');
+  if (!text) return Object.freeze([]);
+  const found = new Map();
+  for (const station of TASHKENT_METRO) {
+    const index = stationMentionIndex(text, station);
+    if (index !== -1) found.set(station.name, { station, index });
+  }
+  for (const [canonical, aliases] of Object.entries(EXTRA_METRO_ALIASES)) {
+    if (found.has(canonical)) continue;
+    const index = extraAliasMentionIndex(text, aliases);
+    const station = TASHKENT_METRO.find((entry) => entry.name === canonical);
+    if (index !== -1 && station) found.set(canonical, { station, index });
+  }
+  return Object.freeze([...found.values()].sort((a, b) => a.index - b.index).map((item) => item.station));
+}
+
 const TASHKENT_HOUSING_TRANSIT = Object.freeze([
   locationEntry('Tashkent North Railway Station', 'railway', [
     'Ташкент Северный вокзал',