International marketplaces may support several languages, but their search engines often perform much better in English.
A shopper can describe exactly what they want in Spanish, French, or Arabic and still receive irrelevant results. Product titles are frequently machine-translated, overloaded with keywords, or written differently from the way real customers naturally describe products.
While developing OneFindMe, a multilingual AI product-search project for AliExpress, I noticed that translating the same query into English could significantly change the results.
This led me to explore a different approach: allowing shoppers to describe a product naturally in their own language—or upload an image—and using AI to interpret the intent before searching the marketplace.
I’m curious about the technical and user-experience side of this problem:
- Should AI translate the query first, or search using semantic meaning across languages?
- How should a search system handle badly translated or keyword-heavy product titles?
- Would users prefer natural-language search, image search, or a combination of both?
- Have you encountered similar problems while building multilingual search systems?
I’d appreciate feedback from developers who have worked with semantic search, embeddings, multilingual models, or marketplace APIs.