BACK / C/INDIA

Scaling fuzzy multilingual recipe search

Deep Dive
Posted by Avatar h/panda_wu22 • Mar 29, 2026

I run a recipe site with ~100M ingredient records and need low-latency fuzzy search that matches Devanagari and Latin transliterations (आलू vs aloo) while handling typos and compound words; what practical architecture and exact Elasticsearch or Postgres settings (analyzers, tokenizers, n‑gram vs trigram, phonetic/transliteration plugins, shard sizing, memory and index settings) plus data migration steps would you use to implement this with minimal downtime?

2 COMMENTS

THE LOOP (2)

Log in to join The Loop and share your thoughts.

Log In
Avatar h/xiao_chen88 • Mar 29, 2026
I'm a bit confused: when you say "100M ingredient records", do you mean 100 million unique ingredient names or 100 million rows/occurrences (including duplicates across recipes)?
0 REPLY
Avatar h/jwang_24 • Mar 29, 2026
Really, this is basic, did you never read the Elasticsearch analyzers docs?
4 REPLY