Unconference Session: A Neat Trick for Fuzzy Search
summary
Indexed 100 sports-product records from a demo dataset using Python and a free Ivan free tier OpenSearch cluster, then enables fuzzy matching with Damerau-Levenshtein auto fuzziness to recover typos (e.g., Yanex → Yonex). Query behavior is inspected with the OpenSearch profile API, revealing unwanted numeric drift (Master 1000 matching 1500/5000). Fix converts digits to spelled words and inserts underscores via a mapping character filter, then adds an extra underscore to break close edit distances — fuzzy works but needs careful token mapping.
AI-written summary · about
chapters (14)
- 0:00Introduction
- 0:21Fuzzy search essentials
- 1:03Damerau-Levenshtein explained
- 1:24OpenSearch setup sports data
- 2:29Basic search Dev Tools
- 3:11Handling typos in queries
- 3:53Query expansion Profile API
- 5:17Fuzzy search on numbers problem
- 6:40Fixing number mismatches
- 7:21Text analysis indexed terms
- 8:23Testing number solution
- 8:44Edge case with number nine
- 9:28Key takeaways fuzzy search
- 10:30Q&A
related talks
28mSuper Search with OpenSearch and Python
33mFuzzy Search Algorithms How and When to Use Them
46mFull-Text Search Internals with Elasticsearch
57mUnlocking the Power of Hybrid Search: A Deep Dive into Python-Powered Precision
42mYou don't need Elasticsearch! Fuzzy Search with PostgreSQL and Spring Data
39mOpenSearch Helps You to Adopt a Friend
24mSearch From the Ground Up
28mA Deep Dive into Python-Powered Precision and Scalability