Smart Autocompl ...
summary
Shows a concrete system that replaces a single concatenated Solr fuzzy index with a pre-query matrix that tests each input token against specific fields, implemented as a Scala Solr request handler and AJAX front end. Uses production logs, runs ~50–100 tiny field-queries per request over ~10 million address/POI points, achieves typeahead within 40 ms latency, and improves precision by preferring matches where tokens map to city/street/house-number fields. Calls this technique field inspection and reports clear production improvement.
AI-written summary · about
chapters (11)
- 0:00Introduction to talk
- 1:07About speaker and project
- 2:11Public transport site use case
- 3:15Naive auto-completion demo
- 5:04Data complexity and challenges
- 8:18Limitations of naive approach
- 10:28Field inspection technique
- 13:40Implementation and tuning
- 15:25Results and feedback
- 16:48Conclusion and takeaways
- 17:10Q&A session
related talks
40mSmarter Search with Spark, Solr and Machine Learning
22mSmarter Search With Spark-Solr
52mSerg Masís | AI Track | Compass AI & Data Summit 2025
21mFrom boolean towards semantic ...
37mBuilding advanced RAGs with Elasticsearch OpenSearch
37mFrom Inverted Index to Columnar Vectorized Execution Search
35mYou've got questions. We've got answers! #bbuzz
34mBuilding a WebScale Search Engine with a Data Driven Architecture
