Multilingual Shopping Queries to Marketplace Queries (OneFindMe, 2026)

<p>941 pairs mapping how shoppers phrase a product search <b>in their own language</b> to the short English query that marketplace listings (AliExpress) are indexed under. The target side is the <b>retrieval</b> form — short, noun-first, in the words sellers use — not a literal translation.</p><p>Useful for training or evaluating query rewriting / query translation for e-commerce retrieval, studying how shopping intent differs from catalogue language, and low-resource shopping vocabulary (especially Hebrew).</p><p><b>Fields:</b> source_query, language (ISO 639-1), marketplace_query, origin (curated / learned / observed), times_searched, spoken (1 if seen via voice search).</p><p><b>Privacy:</b> query text only — no timestamps, countries, IP addresses, user or session identifiers; rows with e-mails, URLs, phone-like or long digit strings removed.</p><p><b>Limitations:</b> heavily Hebrew (862 of 941); some mappings are intent-oriented rather than literal; reflects one marketplace's vocabulary in 2026.</p><p>Source: OneFindMe (onefindme.com/), a free multilingual AI product search engine for AliExpress with voice and image search.

Also deposited at Zenodo (doi:10.5281/zenodo.22930843) and Hugging Face.</p>