Upgrade to Pro
— share decks privately, control downloads, hide ads and more …
Speaker Deck
Features
Speaker Deck
PRO
Sign in
Sign up for free
Search
Search
Context Driven Search Ranking and Faceting
Search
Torsten Bøgh Köster
September 14, 2017
Technology
1.5k
2
Share
Embed
Copy iframe code
Copy JS code
Copy link
Start on current slide
Context Driven Search Ranking and Faceting
Torsten Bøgh Köster
September 14, 2017
More Decks by Torsten Bøgh Köster
See All by Torsten Bøgh Köster
LLMs im Griff: Observability, Tracing und Security
tboeghk
0
32
LLMs im Griff: Observability, Tracing und Security
tboeghk
0
49
Oder mache ich es lieber selbst? Wie sich Kosten und Geopolitik auf Cloud-Betrieb auswirken
tboeghk
0
60
Taking an abandoned Solr search from zero to GenAI hero
tboeghk
0
60
Oder mache ich es lieber selbst? Wie sich Kosten und Geopolitik auf Cloud-Betrieb auswirken
tboeghk
0
56
🔪 How we cut our AWS costs in half
tboeghk
0
400
Shared Nothing Logging Infrastructure
tboeghk
0
130
Beyond Cloud: A road trip into AWS and back to bare metal
tboeghk
1
120
Shared Nothing Logging Infrastructure
tboeghk
0
1.4k
Other Decks in Technology
See All in Technology
AIエージェントとPhysical AIが拓く製造業の変革(ハノーバーメッセリキャップ)
iotcomjpadmin
0
190
Docker Desktop不要の時代が来る? WSL標準の「wslc」で Linuxコンテナを動かしてみた.
ueponx
0
330
製造現場での生成AIの活用、およびエージェントAIの実装のあり方、AVEVAの取り組み
iotcomjpadmin
0
200
小さいから、全部わかる。— 常駐AI "xangi" のすすめ
sugupoko
0
190
なぜ人は自分のプロジェクトを 「なんちゃってアジャイル」と 自嘲するのか
kozotaira
0
210
Hatena Engineer Seminar 37 jj1uzh
jj1uzh
0
310
4人目のSREはAgent
tanimuyk
0
320
ご挨拶「10周年を迎える共創ラボのこれまでとこれから」
iotcomjpadmin
0
160
PostgreSQL 19 新機能概要 OSC Hokkaido 2026
nori_shinoda
0
260
#エンジニアBooks 30分でわかる 「技術記事を書く技術」 / engineer-books 2026-06-30
jnchito
1
150
Agentic AI 時代のテスト手法: Kiro とはじめるプロパティベーステスト (AWS Summit Japan 2026 | DEV212)
ymhiroki
0
130
スタートアップにAmazon EKSは早すぎる? マルチプロダクト戦略を加速する Platform Engineeringの実践 / Is Amazon EKS Too Soon for Startups? Practical Platform Engineering to Accelerate a Multi-Product Strategy
elmodev09
1
2k
Featured
See All Featured
How to optimise 3,500 product descriptions for ecommerce in one day using ChatGPT
katarinadahlin
PRO
1
3.6k
Site-Speed That Sticks
csswizardry
13
1.2k
WENDY [Excerpt]
tessaabrams
11
38k
Responsive Adventures: Dirty Tricks From The Dark Corners of Front-End
smashingmag
254
22k
GitHub's CSS Performance
jonrohan
1033
470k
How to Get Subject Matter Experts Bought In and Actively Contributing to SEO & PR Initiatives.
livdayseo
0
150
Building an army of robots
kneath
306
46k
The untapped power of vector embeddings
frankvandijk
2
1.8k
Designing for Timeless Needs
cassininazir
1
260
Side Projects
sachag
455
43k
Darren the Foodie - Storyboard
khoart
PRO
3
3.4k
Thoughts on Productivity
jonyablonski
76
5.2k
Transcript
Context Driven Search Ranking and Faceting Torsten Bøgh Köster CTO,
shopping24 internet group
@tboeghk #cto, Search Technology Meetup #hamburg, #10YearsSolr, #roadbike, #landscapephotography Torsten
Bøgh Köster |
[email protected]
| @tboeghk | developer.s24.com
e-commerce product search
related products on 3rd party websites
Open Source Power Delivered.
10 years live on Apache Solr
highly customized solr
solr.cool extension directory
Search is contextless
Make search context aware
personalization
conversational context @chatshopper
category context
upfront decision for known queries query —> category
ambiguous queries „braun“
What about new, trending queries?
who wants to configure ∞ query entries?
(1) Detect category context through current search result inspection
Define contexts as Solr queries
represent category context definition as bitset Load JSON from Zookeeper
Evaluate to BitSet in newSearcher event
BitSets are fun! docid —> boolean
testing current query against category context definitons
Build a PostFilter … QParserPlugin … Collector
decisions for a full search result
handle collisions
(2) faceting depending on detected context
Configure facets per context
Add own component to configure faceting
(3) apply ranking depending on detected context
Change scoring during scoring by breaking deferring delegation
None
Gets called for each docid and returns the scroe
None
(4) Make it fast
Introduce a bounded ranking buffer
Fill a bounded buffer. Decide context upon full buffer or
finish() call
Next: Make it stable, Make it better
None
Image Credits https://www.flickr.com/photos/salihan/7040163843 https://www.flickr.com/photos/scrapthispack/2034500846 https://www.flickr.com/photos/135652812@N05/23375030520 https://www.flickr.com/photos/cozymax/10345521973 https://www.flickr.com/photos/juliegallo/9503466665 https://www.flickr.com/photos/platform3/11350560215 https://www.flickr.com/photos/photo64/6341135231 https://www.flickr.com/photos/opoterser/3760102198
https://www.flickr.com/photos/smemon/14931786713 https://www.flickr.com/photos/cyclotourist/25362023650 https://www.flickr.com/photos/lain32/8321828310 https://www.flickr.com/photos/oter/5238799883 https://www.flickr.com/photos/clement127/16056864658 https://www.flickr.com/photos/brendanlynch/4765970880 https://www.flickr.com/photos/rrenomeron/8771565313/ https://www.flickr.com/photos/eugenuity/34113551603 https://www.flickr.com/photos/justinwkern/4535555236