RealEstate Logo

Semantic Search QA Test Suite

Multilingual synonym-based search validation for short-term property discovery

Search Architecture
Synonym-based Semantic Search
1

Synonym Mapping

SEMANTIC_DICTIONARY maps intent concepts β†’ multilingual aliases (EN + PT). Not a lexical keyword search.

2

Fuzzy Matching

Levenshtein distance ≀ 1 for typo tolerance on words β‰₯ 4 characters.

3

Content-Scan Fallback

Scans all 5 language variants (en, es, fr, ar, pt) of title, description, location, and amenities.

4

Deterministic Override

Synonym layer runs BEFORE the LLM and always wins β€” no reliance on LLM for classification.

Test Cases (10)
#QueryLangVariationStatuspropertyTypeKeywords Found
1short termEnglishExact matchβ€”β€”β€”
2shorttermEnglishCompound (no space)β€”β€”β€”
3curta duraΓ§Γ£oPortugueseExact with accentβ€”β€”β€”
4aluguer curta duracaoPortugueseWithout accentβ€”β€”β€”
5vacation rentalEnglishSynonymβ€”β€”β€”
6alojamento localPortugueseSynonym (local lodging)β€”β€”β€”
7shrt termEnglishTypo (missing β€œo”)β€”β€”β€”
8airbnbEnglishBrand / slangβ€”β€”β€”
9estadia curtaPortugueseSynonym (short stay)β€”β€”β€”
10temporadaPortugueseSingle word (season)β€”β€”β€”
Debugging Checklist