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

cookpad で学ぶ自然言語処理/internship2015

j.harashima
August 26, 2015
35k

cookpad で学ぶ自然言語処理/internship2015

Summer Internship 2015 で使用した資料です。

https://recruit.cookpad.com/internship/summer2015-engineer/

j.harashima

August 26, 2015
Tweet

More Decks by j.harashima

Transcript

  1. DPPLQBEͰֶͿࣗવݴޠॲཧ
    ݪౡ७

    View Slide

  2. ࣗݾ঺հ
    ɾݪౡ७
    ɹɾॴଐɿݕࡧฤ੒෦ɺݚڀ։ൃνʔϜ
    ɹɾݞॻɿΤϯδχΞɺ৘ใֶത࢜
    ɹɾઐ໳ɿࣗવݴޠॲཧ

    View Slide

  3. ߨٛɿࣗવݴޠॲཧΛͬ͘͟Γֶ΅͏

    View Slide

  4. ࣗવݴޠॲཧ
    ࣗવݴޠΛίϯϐϡʔλͰॲཧ͢ΔͨΊͷֶ໰෼໺
    ࣗવݴޠࣗવൃੜతʹੜ·ΕͨݴޠʢFH೔ຊޠʣ
    ਓ޻ݴޠਓ޻తʹ࡞ΒΕͨݴޠʢFH$ݴޠʣ

    View Slide

  5. ࣗવݴޠॲཧʹجͮ͘ΞϓϦέʔγϣϯ
    ɾݕࡧΤϯδϯ
    ɹɾ(PPHMFɺ:BIPPɺ
    ɾࣗಈ຋༁
    ɹɾ(PPHMF຋༁ɺΤΩαΠτ຋༁ɺ
    ɾ࣭໰Ԡ౴
    ɹɾ8BUTPOɺ4JSJɺ
    ɾԾ໊׽ࣈม׵
    ɹɾ(PPHMF*.&ɺ"50,ɺ

    View Slide

  6. جૅղੳ
    ܗଶૉղੳ
    ߏจղੳ
    ݻ༗දݱೝࣝ
    ֨ղੳ
    লུরԠղੳ
    ಉٛදݱೝࣝ
    ޠٛᐆດੑղফ
    ஊ࿩ߏ଄ղੳ
    ୯ޠϨϕϧ
    ߏจϨϕϧ
    จ຺Ϩϕϧ
    ҙຯϨϕϧ
    ΞϓϦέʔγϣϯ

    View Slide

  7. ࠓ೔ͷτϐοΫ
    ɾجૅղੳ
    ɹɾܗଶૉղੳ
    ɹɾݻ༗දݱೝࣝ
    ɹɾߏจղੳ
    ɹɾলུরԠղੳ
    ɹɾஊ࿩ߏ଄ղੳ
    ɹɾ֨ղੳ
    ɹɾಉٛදݱೝࣝ
    ɹɾޠٛᐆດੑղফ
    ɾΞϓϦέʔγϣϯ
    ɹɾݕࡧΤϯδϯ
    ɹɾࣗಈ຋༁
    ɹɾ࣭໰Ԡ౴
    ɹɾ͔ͳ׽ࣈม׵

    View Slide

  8. ܗଶૉղੳ
    ୯ޠͷ۠੾Γͱ඼ࢺɺݪܗΛࣝผ͢Δॲཧ
    ଠ࿠ ͸ ژ౎ ʹ ߦͬͨ
    ଠ࿠͸ژ౎ʹߦͬͨ
    ໊ࢺ ໊ࢺ ಈࢺ
    ॿࢺ ॿࢺ
    c c c c
    ߦ͘
    ղੳ༻ͷࣙॻΛࢀরͯ͠ɺҰ൪΋ͬͱ΋Β͍͠
    ʢҰ൪ίετͷ૯࿨͕௿͍ʣ୯ޠྻΛબ୒͢Δ
    ଠ࿠ ໊ࢺ
    ژ౎ ໊ࢺ
    ژ ໊ࢺ
    ౎ ໊ࢺ
    ࣙॻ
    ੜىίετ
    ੜىίετͷ΄͔ʹ୯ޠؒͷ࿈઀ίετ΋ߟྀ

    View Slide

  9. ܗଶૉղੳJODPPLQBE
    ಲͷʹΜʹ͘ে༉ম͖
    ͷ ʹΜʹ͘ ে༉
    ໊ࢺ ໊ࢺ
    ॿࢺ ໊ࢺ
    c c c c
    ম͘
    ಲ ম͖
    ಈࢺ

    View Slide

  10. ݻ༗දݱೝࣝ
    จதͷݻ༗දݱʢFHਓ໊ɺ஍໊ʣΛೝࣝ͢Δॲཧ
    ଠ࿠ ͸ ژ౎ ʹ ߦͬͨ
    ໊ࢺ ໊ࢺ ಈࢺ
    ॿࢺ ॿࢺ
    c c c c
    ߦ͘
    ଠ࿠ ͸ ژ౎ ʹ ߦͬͨ
    ਓ໊ ஍໊
    ܇࿅σʔλΛ༻ҙͯ͠ɺͲͷΑ͏ͳ୯ޠྻ͕
    ͲͷΑ͏ͳݻ༗දݱʹͳΓ͏Δ͔Λֶश
    ࣍࿠͸ژ౎ʹߦͬͨɻ
    Ֆࢠ͸େࡕʹߦͬͨɻ
    ܇࿅σʔλ
    ਓ໊
    ਓ໊
    ஍໊
    ஍໊
    ݻ༗දݱೝࣝث

    View Slide

  11. ಛച৘ใʢళฮ͔Β౤ߘʣ
    ݻ༗දݱೝࣝ
    ͨ·Ͷ͗ݸ ԁ
    ͨ·Ͷ͗υϨογϯά ԁ

    ͨ·Ͷ͗ɹݸ
    ঎඼໊
    ͨ·Ͷ͗ɹυϨογϯά
    ঎඼໊
    ʮͨ·Ͷ͗ʯͰొ࿥
    ʮͨ·Ͷ͗υϨογϯάʯͰొ࿥
    0OMJOF
    Ϣʔβͷۙลͷళฮ͔Β
    ঎඼໊ʮͨ·Ͷ͗ʯΛݕࡧ
    ݻ༗දݱೝࣝJODPPLQBE

    View Slide

  12. ߏจղੳ
    จͷߏ଄ʢ೔ຊޠͰ͸จઅͷ۠੾Γͱ܎ΓઌʣΛ໌Β͔ʹ͢Δॲཧ
    ҎԼͷ৘ใ͔Βࣝผ
    ɾจઅ͸લ͔ΒޙΖʹ܎Δ
    ɾจઅ͸ͨͩҰͭͷจઅʹ܎Δ
    ɾ܎Γड͚͸ޓ͍ʹަࠩ͠ͳ͍
    ɾ࠷΋͍ۙจઅʹ܎Γ΍͍͢
    ɾ
    ଠ࿠ ͸ ژ౎ ʹ ߦͬͨ
    ໊ࢺ ໊ࢺ ಈࢺ
    ॿࢺ ॿࢺ
    c c c c
    ߦ͘
    ଠ࿠͸ ژ౎ʹ ߦͬͨ
    c c
    จઅͷ۠੾ΓΛࣝผ
    ߦͬͨ
    ଠ࿠͸ ژ౎ʹ
    จઅͷ܎ΓઌΛࣝผ
    จઅ͸ݸҎ্ͷ಺༰ޠʢFH໊ࢺʣͱݸ
    Ҏ্ͷ෇ଐޠʢFHॿࢺʣ͔ΒͳΔ

    View Slide

  13. লུরԠղੳ
    রԠࢺ
    จষதͷಉҰදݱΛಉఆ͢Δॲཧ
    ઌߦࢺ
    রԠղੳ লུղੳʢθϩরԠղੳʣ
    ޲͔ͬͨ
    ൴͸ ਗ਼ਫࣉʹ
    ·ͣ
    ଠ࿠͸ ژ౎ʹ ߦͬͨ
    θϩ୅໊ࢺ
    ઌߦࢺ
    ޲͔ͬͨ
    ൴͸ ਗ਼ਫࣉʹ
    ·ͣ
    ଠ࿠͸ ژ౎ʹ ߦͬͨ
    θϩ୅໊ࢺʢলུ͞ΕͨরԠࢺʣʹؾ෇͘ͱ͜Ζ͔Β
    ελʔτ
    ҎԼͷ৘ใ͔Βಉఆ
    ɾরԠࢺͱઌߦࢺͷੑผ͸ಉ͡
    ɾઌߦࢺ͸ʮ͸ʯΛ൐͍΍͍͢
    ɾরԠࢺͱઌߦࢺ͸ͦΕ΄Ͳ཭Εͳ͍
    ɾ

    View Slide

  14. ߏจղੳBOEলུরԠղੳJODPPLQBE
    ɾࠓͷͱ͜Ζɺ࢖͍Ͳ͜Ζͳ͠
    ɹɾࣗಈ຋༁ʢޙͰ঺հʣͰ࢖༻͢Δ͔΋ʁ

    View Slide

  15. ୯ޠϨϕϧ ߏจϨϕϧ
    ίϯϐϡʔλɹʹɹܭࢉػ
    ίϯϐϡʔλɹʹɹίϯϐϡʔλʔ ܠؾ͕ྫྷ͑Δɹʹɹܠؾ͕Լ͕Δ
    ಉٛදݱೝࣝ
    ಉ͡ҙຯΛ࣋ͭදݱΛೝࣝ͢Δॲཧ
    ্͸ҟදهಉٛʢಉ͡୯ޠͷදه༳Εʣ
    Լ͸ҟܗಉٛʢҟͳΔ୯ޠ͕ಉ͡ҙຯʣ
    ୯ޠϨϕϧͰ͸ʮྫྷ͑Δʯ㱠ʮԼ͕Δʯ͕ͩ
    ʮܠؾ͕ʯ͕෇͘ͱಉٛ

    View Slide

  16. ಉٛදݱೝࣝJODPPLQBE

    View Slide

  17. ݕࡧΤϯδϯ
    ΫΤϦʹؔ࿈͢ΔϦιʔεʢFHจॻʣΛऔಘ͢Δॲཧ
    ΠϯσΩγϯά
    సஔΠϯσοΫε
    ୯ޠ͔Βจॻ΁ͷϚοϐϯά
    0⒐JOF 0OMJOF
    จॻଠ࿠͸ژ౎ʹߦͬͨ
    จॻ࣍࿠͸ژ౎ʹߦͬͨ
    ܗଶૉղੳ
    จॻଠ࿠c͸cژ౎cʹcߦͬͨ
    จॻ࣍࿠c͸cژ౎cʹcߦͬͨ
    ଠ࿠\^
    ࣍࿠\^

    ژ౎\ ^
    ΫΤϦʮଠ࿠ʯ
    ݕࡧ݁Ռɿจॻ
    ΫΤϦʮژ౎ʯ
    ݕࡧ݁Ռɿจॻ จॻ

    View Slide

  18. ݕࡧΤϯδϯJODPPLQBE

    View Slide

  19. ࣗಈ຋༁
    ίϯϐϡʔλʹΑΔผͷݴޠͷ౳Ձͳจষ΁ͷஔ׵
    ೔ɿଠ࿠͸ژ౎ʹߦͬͨ
    ӳɿ5BSPXFOUUP,ZPUP
    ܗଶૉղੳɺߏจղੳɺ
    ݴޠ୯Ґ͝ͱʹ຋༁
    ߦͬͨ
    ଠ࿠͸ ژ౎ʹ
    5BSP UP,ZPUP XFOU MFBWF
    ࣗવจΛੜ੒
    ೖྗจΛಛఆͷݴޠ୯ҐʢFH୯ޠɺ୯ޠྻʣ
    Ͱදݱ
    ୯Ґ͝ͱͷީิΛ૊Έ߹Θͤͯෳ਺ͷจΛੜ੒
    ͦͷத͔Β΋ͬͱ΋ࣗવͳ΋ͷΛग़ྗ
    ର༁σʔλΛࢀߟʹ຋༁ީิΛੜ੒ʢσʔλ͸
    ༧Ίߏங͓ͯ͘͠ʣ

    View Slide

  20. ࣗಈ຋༁JODPPLQBE
    8FBSFIJSJOH

    View Slide

  21. ࣮शɿܗଶૉղੳثΛ࢖ͬͯΈΑ͏

    View Slide

  22. ܗଶૉղੳث




    +6."/
    $IB4FO
    .F$BC
    ,Z5FB
    ػցֶशʢੜ੒ϞσϧʣʹΑΔίετௐ੔
    ػցֶशʢࣝผϞσϧʣʹΑΔίετௐ੔
    ֶशσʔλͷߏஙΛ༰қʹ
    ਓखʹΑΔίετௐ੔

    View Slide

  23. .F$BC
    ɾݱࡏͷσϑΝΫτ
    ɹɾߴਫ਼౓Ͱ࢖͍΍͍͢
    ɾ࢖͍ํ
    ɹɾ45%*/ʹͭͬ͜Ή͚ͩ
    NFDBC
    ଠ࿠͸ژ౎ʹߦͬͨ&OUFS
    ଠ࿠ɹ໊ࢺ ݻ༗໊ࢺ ਓ໊ ໊ ଠ࿠ λϩ΢ λϩʔ
    ͸ɹɹॿࢺ ܎ॿࢺ ͸ ϋ ϫ
    ژ౎ɹ໊ࢺ ݻ༗໊ࢺ ஍Ҭ Ұൠ ژ౎ Ωϣ΢τ Ωϣʔτ
    ʹɹɹॿࢺ ֨ॿࢺ Ұൠ ʹ χ χ
    ߦͬɹಈࢺ ཱࣗ ޒஈɾΧߦଅԻศ ࿈༻λ઀ଓ ߦ͘ Πο Πο
    ͨɹɹॿಈࢺ ಛघɾλ جຊܗ ͨ λ λ
    &04
    ୯ޠͷ۠੾Γͱ඼ࢺɺݪܗ͕෼͔Δ
    ؆୯ͳ΋ͷͰ͋Ε͹ɺݻ༗දݱ΋෼͔Δ

    View Slide

  24. ࣮शϨγϐΛղੳ͠Α͏
    Ϩγϐσʔλʢ ϨγϐͷλΠτϧʣΛऔಘ
    ɹɾDVSMPSFDJQFUYUlIUUQTYYYzʢ63-͸4MBDLࢀরʣ
    ܗଶૉղੳΛ࣮ߦ
    ɹɾNFDBCSFDJQFUYUʁ

    View Slide

  25. ࣮श୯ޠͷ෼෍Λௐࠪ͠Α͏
    ग़ݱස౓Λܭ਺
    ɹɾࢥ͍ࢥ͍ͷํ๏Ͱʂ
    ॱҐ ୯ޠ ग़ݱස౓


    ʁ
    ʁ
    ؆୯
    ̇

    Ϩγϐʹ ճҎ্
    ग़ݱ͢Δ୯ޠͱ͸ʁ

    View Slide

  26. ࣮शδοϓͷ๏ଇΛ؍ଌ͠Α͏
    ॱҐͱग़ݱ֬཰ΛάϥϑԽ
    ɹɾࢥ͍ࢥ͍ͷํ๏Ͱʂ
    શମͰL൪໨ͷग़ݱස౓Λ࣋ͭཁૉ͕શମʹ
    ઎ΊΔׂ߹ʢJFग़ݱ֬཰ʣ͸Lʹൺྫ

    ݱ
    ֬

    ॱҐ
    ʁ

    View Slide

  27. ࣮श୯ޠΛొ࿥͠Α͏
    ܗଶૉղੳΛ࣮ߦ
    ɹɾFDIPlΫοΫύουzcNFDBCʁ
    ࣗ෼༻ͷࣙॻΛ࡞੒
    ɹɾ୯ޠͷ৘ใΛΧϯϚ۠੾ΓͰొ࿥
    ࣙॻΛίϯύΠϧ
    ɹɾVTSMPDBM$FMMBSNFDBCMJCFYFDNFDBCNFDBCEJDUJOEFYE
    ɹɹVTSMPDBM$FMMBSNFDBCMJCNFDBCEJDJQBEJDVVTFSEJDVTFSDTWGVUGUVUG
    ܗଶૉղੳΛ࣮ߦ
    ɹɾFDIPlΫοΫύουzcNFDBCVVTFSEJDʁ
    ද૚ܗ ݪܗ ಡΈ ൃԻ
    ίετ ඼ࢺ
    ΫοΫύου ໊ࢺ ݻ༗໊ࢺ ૊৫ ΫοΫύου ΫοΫύου ΫοΫύου
    VTFSDTW

    View Slide

  28. ಠशɿࣗવݴޠॲཧͱ஥ྑ͘ͳΖ͏

    View Slide

  29. ؔ࿈ਤॻˍؔ࿈ֶձ
    ɾؔ࿈ਤॻ
    ɹɾࣗવݴޠॲཧʢ์ૹେֶڭҭৼڵձʣ
    ɹɾࣗવݴޠॲཧͷجૅʢίϩφࣾʣ
    ɹɾݴޠॲཧͷͨΊͷػցֶशೖ໳ʢίϩφࣾʣ
    ɾؔ࿈ֶձ
    ɹɾݴޠॲཧֶձ
    ɹɾࣗવݴޠॲཧݚڀձ
    ɹɾ/-1एखͷձ

    View Slide