Mirror, Mirror - What am I Typing Next? All About Auto Suggest
summary
Shows concrete Java implementations: a naive linear scan, a radix tree and pruning radix tree (each node stores max child score so lookups prune early), an adaptive radix compression, and Lucene’s weighted FST (build-time, not live-updatable, but serializable — tested millions of titles to ~42 MB). Uses concurrent-trees library (lock free reads, unmaintained), SimSpell and fuzzy suggest for typo tolerance, infix vs prefix behavior, and notes learning‑to‑rank + telemetry are needed for meaningful ordering and sub‑100ms round‑trip performance.
AI-written summary · about
chapters (14)
- 0:00Introduction to Auto Suggest
- 1:05What is Auto Suggest
- 5:16Basic Auto Suggest Implementation
- 7:44Tree-Based Suggest Data Structures
- 10:35Radix and Adaptive Radix Trees
- 13:30Pruning Radix Trees for Relevancy
- 15:55Lucene and FST for Suggestions
- 20:16Fuzzy and Phonetic Search
- 24:09Infix vs Prefix Suggestions
- 26:17Typo Tolerance Techniques
- 28:28Ranking and Relevancy Scoring
- 31:57Learning to Rank in Suggestions
- 33:23Practical Implementation Tips
- 35:50Summary and Future of Auto Suggest
related talks
50mElasticsearch Suggesters: Beyond Autocomplete
32m"You may also like"
13mI got the computer to find words with good anagrams and....!!
22mImage Retrieval using Short Texts
39mMachine Learning for Relevance and Serendipity
16mVisual statical analysis tool
01h 48mRecommendation engines explained
38mRecommendation Engine with Neo4j, Piece of Cake