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
俺の全文検索エンジン(Go製)を作り始めた
Search
kotaroooo0
November 11, 2020
Programming
140
0
Share
Embed
Copy iframe code
Copy JS code
Copy link
Start on current slide
俺の全文検索エンジン(Go製)を作り始めた
kotaroooo0
November 11, 2020
More Decks by kotaroooo0
See All by kotaroooo0
データ鮮度を落とさずに安全にReindexしたい
kotaroooo0
0
110
検索エンジン自作入門 Go Conference 2021 Spring
kotaroooo0
17
7.6k
転置インデックスでどう検索しているか
kotaroooo0
0
380
ぼくのかんがえたさいきょうのDocker Build
kotaroooo0
0
110
Other Decks in Programming
See All in Programming
광주소프트웨어마이스터고등학교 DevFest 특강 - 바이브 코딩 시대에서 주니어 개발자로 살아남는 방법
utilforever
1
150
Android CLI
fornewid
0
120
Claude Team Plan導入・ガイド
tk3fftk
0
220
なぜ関数型プログラミングで「型」と「証明」が語られるのか #fp_matsuri
kajitack
3
1k
ルールを書いて終わらせないハーネスエンジニアリング
yug1224
4
1.7k
ビデオ通話が繋がる0.2秒で何が起きているのか
supurazako
2
150
Embedded SREと共に達成した会員管理システムのAWS移行 - SRE NEXT 2026 ランチスポンサーセッション
niftycorp
PRO
1
2.9k
関数型プログラミングのメリットって何だろう?
wanko_it
0
190
エンジニア向け会社紹介/Findy Company Profile
findyinc
6
360k
ここ半年くらいでAIに作らせたR用ツール
eitsupi
0
120
GitHubCopilotCLIのスラッシュコマンドを自作してみる
htkym
0
100
Augmenting AI with the Power of Jakarta EE
ivargrimstad
0
160
Featured
See All Featured
KATA
mclloyd
PRO
35
15k
Leveraging Curiosity to Care for An Aging Population
cassininazir
1
420
Making the Leap to Tech Lead
cromwellryan
135
10k
Kristin Tynski - Automating Marketing Tasks With AI
techseoconnect
PRO
0
410
How to optimise 3,500 product descriptions for ecommerce in one day using ChatGPT
katarinadahlin
PRO
1
3.7k
Cheating the UX When There Is Nothing More to Optimize - PixelPioneers
stephaniewalter
287
14k
What Being in a Rock Band Can Teach Us About Real World SEO
427marketing
0
1.1k
Making Projects Easy
brettharned
120
6.7k
The untapped power of vector embeddings
frankvandijk
2
1.8k
It's Worth the Effort
3n
188
29k
Google's AI Overviews - The New Search
badams
0
1.1k
Jess Joyce - The Pitfalls of Following Frameworks
techseoconnect
PRO
1
310
Transcript
2020/11/11 @kotaroooo0 ԶͷશจݕࡧΤϯδϯ(Go) Λ࡞Γ࢝Ίͨ
͜ͷLTΛฉ͘ͱ… 1. ͳΜͱͳ͘શจݕࡧΤϯδϯͷΈ͕͔ Δ 2. GoͰશจݕࡧΤϯδϯΛ࡞Γ࢝ΊΒΕΔ
-શจݕࡧΤϯδϯͷΈΛΔ -ElasticsearchͰΘΕ͍ͯΔApache LuceneΆ͍༷ͷͷΛ࡞Δ -GoΛֶΜͰ͍ΔͷͰԿ͔࡞Γ͍ͨ -Twitter BotΛ࡞͍ͬͯΔ͕ɺͪΐ͏Ͳ͍͍શจݕࡧΤϯδϯ͕ͳ͍ -ܰྔɺ͔ͭॏΈ͖ϨʔϕϯγϡλΠϯڑΛܭࢉͰ͖Δͭ શจݕࡧΤϯδϯΛ࡞Δཧ༝
3Ͱ͔ΔશจݕࡧΤϯδϯ
શจݕࡧͷΈ INDEXING ୯ޠ จॻ have 1,2 pen 1 we 2
Desk 2 จॻ1 “I have a pen.” จॻ2 “We have desk.” CHAR FILTER TOKENIZER TOKEN FILTER Analyzer
શจݕࡧͷΈ SEARCH ୯ޠ จॻ have 1,2 pen 1 we 2
Desk 2 ݕࡧϫʔυ: “pen” จॻ1͕ώοτ CHAR FILTER TOKENIZER TOKEN FILTER Analyzer
ANALYZERͳͥඞཁ? - τʔΫϯׂͯ͘͠ΕΔͨΊ - “I have a pen.” ͜ͷ··ͰసஔΠϯσοΫε Λ࡞Ͱ͖ͳ͍ͷͰɺI,
have, a, penͱτʔΫ ϯׂ͍ͨ͠ - ΫΤϦͷදه༳ΕΛٵऩͨ͠Γ͢ΔͨΊ - “GOD”ͱ͍͏୯ޠΛؚΉυΩϡϝϯτ ɺ”god”Ͱώοτ͢ΔΑ͏ʹখจࣈʹ౷Ұ ͍ͤͨ͞ - ແବͳτʔΫϯͷϑΟϧλϦϯά - theͳͲΠϯσΩγϯάͯ͠ແବ Analyzeલ Analyzeޙ “I have a BIG pen!” have, big, pen
۩ମతͳANALYZERྫ - Char Filter(Tokenizerͷલʹॲཧ͢Δ) 0ݸҎ্ - Mapping: إจࣈΛ୯ޠʹมͳͲ - HTMLstrip:
HTMLΛύʔε - Tokenizer(τʔΫϯׂ͢Δ) 1ݸ - Standard: εϖʔεͳͲϧʔϧʹैׂͬͯ - Kuromoji: ܗଶૉղੳͰׂ - Ngram: Nจࣈ͝ͱʹׂ - Token Filter(Tokenizerͷޙʹॲཧ͢Δ) 0ݸҎ্ - Lowercase: খจࣈ - Stopword: ετοϓϫʔυআڈ - Stemming: දه༳Ε CHAR FILTER TOKENIZER TOKEN FILTER Analyzer
࣮
ANALYZERͷ࣮ߦΠϝʔδ Analyzeલ MappingCharFilter StandardTokenizer LowercaseFilter StopWordFilter StemmerFilter I have a
lot of TASKs. I am very sad :( I have a lot of TASKs. I am very sad _sad_ I, have, a, lot, of, TASKs, I, am, very, sad, sad I, have, a, lot, of, tasks, I, am, very, sad, sad lot, tasks, am, very, sad, sad lot, task, am, very, sad, sad ॲཧͷྲྀΕ
ANALYZERͷ࣮
CHAR FILTERͷ࣮
TOKENIZERͷ࣮
TOKEN FILTERͷ࣮
INDEX - సஔΠϯσοΫε map[string][]int - సஔΠϯσοΫεΛϑΟʔϧυ ໊͝ͱʹ࣋ͭ - υΩϡϝϯτɺIDͱϑΟʔϧ υΛ࣋ͭ
- Indexing͢Δͱ͖AnalyzerΛ ௨͢
SEARCH -ANDݕࡧͱORݕࡧ -ΫΤϦ: “pink orange blue” -ANDݕࡧ: 3 -OR: ݕࡧ1,2,3,4,5,6
pink Orange blue 6 3 4 5 2 1
ಈ͔ͯ͠ΈΔ
ݕࡧྫ - ※Analyzer͖ͬ͞ͱಉ༷ - IndexSearchAnalyzerΛ௨͢ - ”foxes”Ͱ”fox”ΛؚΉυΩϡϝϯτ͕Ϛον - ”happy”Ͱ”:)”ΛؚΉυΩϡϝϯτ͕Ϛον -
ANDݕࡧ - fine,FaX,foxes,happy͕શؚͯ·Ε͍ͯΔυΩϡϝ ϯτ1,2͕ώοτ͍ͯ͠Δ
ࠓޙ -͍͋·͍ݕࡧΛ࣮͢Δ -Fuzzy QueryɺSuggesters -είΞܭࢉΛ࣮͢Δ -IFIDF, BM25
ࢀߟ -https://github.com/kotaroooo0/stalefish -https://artem.krylysov.com/blog/2020/07/28/lets-build-a-full-text-search-engine/