Upgrade to Pro — share decks privately, control downloads, hide ads and more …

speee-kaigi-160629

 speee-kaigi-160629

SpeeeKaigi#1で会社の名寄せにGraphDB(Neo4j)を使ってみた話をした際のスライドです

kawakubox

June 29, 2016
Tweet

More Decks by kawakubox

Other Decks in Programming

Transcript

  1. ࠓճͷ໊دʹ༻͍ͨσʔλϞσϧ Relation • has [CompanyName -> Word] • ձ໊ࣾ ͕

    ୯ޠ Λ͍࣋ͬͯΔؔ܎Λද͢ • similar [Word -> Word] • ୯ޠ ͱ ୯ޠ ͕ࣅ͍ͯΔؔ܎Λද͢ • ϨʔϕϯγϡλΠϯڑ཭͕͔ۙͬͨΓɺྨ ٛޠͩͬͨΓ
  2. ຊ൪؀ڥͷσʔλΛ౤ೖ • Node • CompanyName … 25,947 • Word …

    16,944 • Relation • has … 99,526 • similar … 11,546 • σʔλαΠζ … ͓Αͦ 90MB
  3. GraphDBͰݕࡧͯ͠ΈΔ MATCH (base: CompanyName {text: LOWER('AKR Land Deveploment PT')}) MATCH

    p=(base)-[*1..2]->()<--(c) WHERE SIZE((base)-[*1..2]->()<--(c)) > base.word_count - 1 RETURN DISTINCT c AKR Land Deveploment PT ͱ͍͏typo͍ͯ͠Δ ձ໊ࣾͰݕࡧΛ͔͚Δͱɺɺɺ akr land development, pt akr land development pt ͕ݕࡧ݁Ռͱͯ͠දࣔ͞ΕΔɻ
  4. ΍ͬͯΈͯ • Pros • RDBMSͰ͋Γ͕ͪͳςʔϒϧ࠶ؼ݁߹ͷί ετθϩ • ܦ࿏୳ࡧϥΫ • Cons

    • Cypher ͱ͍͏൚༻Ͱ͖ͳ͍ΫΤϦݴޠ • Wordϊʔυ͕ଟ͍ͷͰ Wordؒͷؔ܎