toC企業でのデータ活用 (PyData.Okinawa + PythonBeginners沖縄合同勉強会 2019)

by takegue

Slide 1

Slide 1 text

PyData.Okinawa + PythonBeginnersԭೄ ߹ಉษڧձ 2019 ஛໺ फ़ี (@takegue) toCاۀͰͷσʔλ׆༻; αΠΤϯεɺΤϯδχΞϦϯάͦͯ͠σβΠϯɺΞʔτ

Slide 2

Slide 2 text

8IP"N* ஛໺ फ़ีʢ @takegue ʣ Retty ← म࢜ʢNLP; ػց຋༁ʣˡ ߴઐ  Core Value: Data Architect   σʔλͷՁ஋Λ࠷େԽ͢Δ࢓૊Έ/ઃܭͷ࣮ݱ    ࣥච׆ಈ:   ʮ༏ઌ౓ֶशʹΑΔਪનจ͔Βͷݟग़͠நग़ʯ  ʮ΍ͬͯΈΑ͏ʂ ػցֶशʢSotware Designʣʯ  ʮࢼֶͯ͠Ϳ ػցֶशೖ໳ʯଞ…    ߴ౓AIਓࡐ͔΋ʁ  ͦͷଞ: https://shwca.se/takegue

Slide 3

Slide 3 text

Data Architectͷ͓͠͝ͱ: ྲྀ௨ͱσʔλͷܦࡁݍΛ࡞Δ͜ͱ ΞφϦετ σʔλϚʔτ ϓϩμΫτ σʔλ΢ΣΞϋ΢ε

Slide 4

Slide 4 text

σʔλʹؔΘΔδϣϒɺ͍ΖΜͳδϣϒ͕͋Δ • Data Scientists • Data Infrastructure Engineer • ML Engineer / SysML • BI Engineer / Data Platform Engineer • Data Visualization Engineer / Data Analyst • Data Application Engineer

Slide 5

Slide 5 text

܅͸ͲΜͳδϣϒʹͳΓ͍ͨʁʂ

Slide 6

Slide 6 text

͜ͷઌੜ͖࢒ΔͨΊʹ͜͏ͳΓ͍ͨ ʮ͜ͷձࣾͷ໋ߝ͸Զ͕Ѳ͍ͬͯΔʯʢը૾ུʣ ݴͬͯΈͨ͘ͳ͍ʁ

Slide 7

Slide 7 text

toCاۀͱͯ͠ͲͷΑ͏ʹσʔλ׆༻ʹऔΓ૊ΜͰ͍Δͷ͔ʁ

Slide 8

Slide 8 text

·ͣ͸αʔϏε঺հ: Retty

Slide 9

Slide 9 text

No content

Slide 10

Slide 10 text

https://retty.me/announce/philosophy/

Slide 11

Slide 11 text

No content

Slide 12

Slide 12 text

σʔλʹؔΘΔλεΫɺ͍ΖΜͳλεΫ͕͋Δ

Slide 13

Slide 13 text

σʔλʹؔΘΔλεΫɺ͍ΖΜͳλεΫ͕͋Δ ΞϓϦ πʔϧ

Slide 14

Slide 14 text

σʔλʹؔΘΔλεΫɺ͍ΖΜͳλεΫ͕͋Δ ΞϓϦ πʔϧ ૊৫ن໛Ͱͷ  εέʔϧϝϦοτ͕ߴ͍ Ϣʔβن໛Ͱͷ  εέʔϧϝϦοτ͕ߴ͍

Slide 15

Slide 15 text

σʔλʹؔΘΔλεΫɺ͍ΖΜͳλεΫ͕͋Δ ந৅త ۩ମత ΞϓϦ πʔϧ

Slide 16

Slide 16 text

σʔλʹؔΘΔλεΫɺ͍ΖΜͳλεΫ͕͋Δ ந৅త ۩ମత ΞϓϦ πʔϧ ௕ظˍܧଓత։ൃ޲͚ʢR&Dʣ ୹ظత/ूதత։ൃ ςί͕ޮ͖΍͍͢;   3ഒͷੜ࢈ੑˠ 10ഒͷ੒ՌʹมΘͬͨΓ͢Δ 10ഒͷੜ࢈ੑͷҧ͍͕ͦͷ··10ഒͷ੒Ռͷࠩ

Slide 17

Slide 17 text

σʔλʹؔΘΔλεΫɺ͍ΖΜͳλεΫ͕͋Δ ந৅త ۩ମత ΞϓϦ πʔϧ Ϩίϝϯυ ݕࡧ ίϯςϯπੜ੒ ଟݴޠରԠ ޿ࠂ ίϯςϯπ؂ࢹ ؂ࢹ (ҟৗݕ஌౳) ऩӹ༧ଌ ࣗಈQA ୳ࡧతσʔλ෼ੳ ϝτϦΫε։ൃ ύϑΥʔϚϯε෼ੳ Ծઆݕূ

Slide 18

Slide 18 text

σʔλʹؔΘΔλεΫɺ͍ΖΜͳλεΫ͕͋Δ ந৅త ۩ମత ΞϓϦ πʔϧ Ϩίϝϯυ ίϯςϯπੜ੒ ଟݴޠରԠ ޿ࠂ ίϯςϯπ؂ࢹ ؂ࢹ (ҟৗݕ஌౳) ऩӹ༧ଌ ࣗಈQA ୳ࡧతσʔλ෼ੳ ϝτϦΫε։ൃ ύϑΥʔϚϯε෼ੳ Ծઆݕূ ݕࡧ

Slide 19

Slide 19 text

ػցֶशϓϩΤΫτͷҰྫ: ʮ༏ઌ౓ֶशʹΑΔਪનจ͔Βͷݟग़͠நग़ʯ  from http://www.orsj.or.jp/archive2/or62-11/or62_11_731.pdf

Slide 20

Slide 20 text

ΩϟονίϐʔΛࣗಈతʹܾΊΕΔΑ͏ʹ͍ͨ͠ ୺తʹ͍͏ͱ  ͓ళͷͨΊͷΩϟονίϐʔΛ࡞Δ

Slide 21

Slide 21 text

͸͡Δͱ͖ʹԿΛߟ͑Δ͔ʁ •ΤϯδχΞϦϯάγϯΩϯά (ٕज़తʹͳΜͱ͔͢Δ) •ϓϩμΫτγϯΩϯά (Ϣʔεέε/Ձ஋ਫ४ΛఆΊΔ) •αΠΤϯεγϯΩϯά (໰୊ͷຊ࣭Λ໰͏)

Slide 22

Slide 22 text

ϓϩμΫτγϯΩϯά • ඼࣭ج४ͱͯ͠ʮ৴པʯΛଛͳΘͳ͍͔ʁ • ӕΛ͔ͭͳ͍͜ͱ • ޱޠతͰ͸ͳ͍ͳͲʮΒ͠͞ʯΛද͢બ޷ੑ͕͋Δ • αʔϏεͷڧΈʹͳΔΑ͏ͳ΋ͷ͕๬·͍͠ • Ωϟονίϐʔͱͯ͠ͷཱͪҐஔ; ັྗతͳจͰ͋Δ͜ͱ • ͋ͨΓ͞ΘΓͷͳ͍ฏۉతͳจষΛٻΊ͍ͯΔΘ͚Ͱ͸ͳ͍  • ϓϩμΫτ΁ͷ౷߹ͷ਌࿨ੑ͕ߴ͍͜ͱ͕๬·͍͠ • จࣈ਺੍ݶͷ໰୊ (PC΍εϚϗ)

Slide 23

Slide 23 text

αΠΤϯεγϯΩϯά • ັྗతʹײ͡Δจͱ͸Կ͔ʁ ◦ධՁͷઃܭ ▪ આಘྗ͕͋ΔΩϟονίϐʔ → CTR্͕Γͦ͏ʁ ▪ CTR͕͕͋ΔΩϟονίϐʔ ≠ આಘྗʁ  • ΩϟονίϐʔΒ͍͠ͱ͸Կ͔ʁ ◦Ωϟονίϐʔͷྲྀெੑ ≠ จͱͯ͠ͷྲྀெੑ ◦จͱͯ͠͸ଟগ่Ε͍ͯͯ΋ྑ͍ʢϦζϜ͕͋Δͱྑ͍ʣ ◦ʮ͜ͷ͓ళͷεύήοςΟ͸ඒຯ͍͠Ͱ͢ʯ ◦ʮඒຯͳεύήοςΟΛఏڙʂʯ • Ωϟονίϐʔ͸ʮޱίϛʯͷཁ໿ͳͷ͔ • ͦ΋ͦ΋NLPͱͯ͠͸Ͳ͜·Ͱ͕Ͱ͖Δൣғͳͷ͔ʁ ◦ ػցతʹྲྀெͳจΛੜ੒͢Δ͜ͱ͸Ͱ͖Δ͔ʁ ▪ ػցֶशόοΫάϥ΢ϯυͱͯ͠ͷ஌ݟ ◦ Ͳ͏͍͏࣮ݧઃఆͩͬͨΒ͏·͘ਐΊΒΕΔ͔ʁ

Slide 24

Slide 24 text

ΤϯδχΞϦϯάʹ͜ΕΒΛ౿·͑ͯͳΜͱ͔͢Δ • ϓϩμΫτΠϯͷखؒ͸ʁ ◦ DBʹಥͬࠐΜͰͪΐͬͱίʔυΛॻ͖׵͑Δ͚ͩɺ͓खܰ؆୯ʂ ◦ ࢼߦࡨޡͷํʹ͕͖࣌ؒ͞΍͍͢ • ࠷ߴਫ਼౓ͷख๏͕ඞཁͳ༁Ͱ͸ͳ͍ ◦ ख๏ࣗମʹ৽نੑ͕ͳͯ͘ྑ͍ɻ஌ݟͷ৽نੑ͸ཉ͍͠ ◦ ࢼͯ͠ධՁͯ͠վળͰ͖Δ΋ͷ͕ྑ͍ ◦ ֶशʹ͕͔͔࣌ؒΔ௒େن໛ֶश͸࠷ॳ͸΍Βͳ͍ • ݱঢ়͋Δσʔληοτͷ೺Ѳ ◦ Ωϟονίϐʔͷจ਺͸͔ͳΓ͋Δ (20ສจڧ) ˍ ޱίϛ΋ͨ͘͞Μ͋Δʂ ◦ ੜ੒͢ΔͨΊʹ׬શʹ੔උ͞Εͨσʔληοτ͸ͳ͍ ˍ ୹ظܾઓ (1.0ϱ݄) ◦

Slide 25

Slide 25 text

ྫ͑͹ … • ςϯϓϨʔτࢤ޲ ◦ ௒େྔͷൈ͚͕݀͋ΔςϯϓϨʔτΛ༻ҙ͠  ٖࣅతʹେྔͷจΛੜ੒͠ɺͦ͜ͷத͔Βྑ͍΋ͷΛબͿ ▪ ΩϟονίϐʔͷݴޠϞσϧͰྲྀெੑ͸ධՁͰ͖Δʂ ▪ ΩϟονίϐʔͷςϯϓϨʔτΛ͍͔ʹఏڙͰ͖Δ͔ʁ • ׬શจੜ੒ࢤ޲ ◦ GANGAN͍͜͏ͥʂ ◦ ΍ͬͨ͜ͱͳ͍͠ɺ΍ָͬͯͯͦ͠͏  • ޱίϛཁ໿ࢤ޲ ◦ ޱίϛΛཁ໿੍ͯ͠ݶ͞ΕͨจࣈͰจΛͭ͘Δ

Slide 26

Slide 26 text

Ͳ͏͔ͨ͠ʁ • ཁ໿ʢநग़ʣࢤ޲ͷΞϓϩʔνͱͯ͠໰୊ΛϞσϧԽ • ̎ͭͷจʹରͯ͠ɺࣄྫؒͷॱংؔ܎>= Λֶश͢Δ2஋෼ྨثΛߏங͢Δ໰୊ͱͯ͠ϞσϧԽ ɹɹɹ   ྑ͍ΩϟονίϐʔΛઈରతͳࢦඪͰܭଌ͢Δͷ͸೉͍͕͠ ɹɹɹ ૬ରతͳؔ܎͸؆୯ʹఆٛͰ͖Δɻ ɹɹɹ ɹɹɹॱংؔ܎͕ఆٛͰ͖Δͱιʔτ͕Ͱ͖Δʂ f(X1 , X2 ) = F(ϕ(X1 ) − ϕ(X2 ))) = { 1, if X1 ≥ X2 0, otherwise f(“͜ͷ͓ళͷຯḩो͸͏·͍”, “ࣗՈ੡ͷຯḩो͸͓;͘Ζͷຯʂ”ʣ  = “͜ͷ͓ళͷຯḩो͸͏·͍” =< “ࣗՈ੡ͷຯḩो͸͓;͘Ζͷຯʂ"

Slide 27

Slide 27 text

Ͳ͏͔ͨ͠ʁ • Ωϟονίϐʔͷจ >= ޱίϛ͔ΒϥϯμϜʹ੾Γग़ͨ͠จɹͰେྔͷٖࣅσʔλΛੜ੒ ɹɹ େྔͷ܇࿅ࣄྫˍग़ྗͷ࣍ݩ਺2Ͱ͋ΔͨΊɺֶशͰ͖ͦ͏ͳؾ͕͢Δ ΦϯϥΠϯߋ৽͕ՄೳͳϩδεςΟοΫճؼΛ෼ྨثʹར༻͢Δ͜ͱͰ  ɹɹ σʔλྔʹରͯ͠΋໰୊ͳֶ͘शͰ͖ΔΑ͏ʹ (sklearn.linear_model.SGDClassiﬁer Λར༻) ɹɹ Ұ؏ੑͷ͋Δσʔλྔ͕े෼ʹ֬อͰ͖Δͱ NNܥͷػցֶश͸ɺ͍͍ͩͨͲΜͳࣸ૾Ͱ΋Ͱ͖Δ ৄࡉ͸ׂѪ (http://www.orsj.or.jp/archive2/or62-11/or62_11_731.pdf)

Slide 28

Slide 28 text

Ͳ͏͔ͨ͠ʁ ◦ ྑ͍ޱίϛ͔Βྑ͍Ωϟονίϐʔ͕͓ళʹ০ΒΕΔʂ • Ϣʔβͷޱίϛ͕͓ళΛԠԉ͢Δͱ͍͏ɺαʔϏεͷՁ஋؍ͱ΋Ϛον ◦ ॊೈੑ͕ߴ͍: ਪ࿦ϑΣʔζͷࡍͷจͷੜ੒ํ๏Λ޻෉͢Ε͹ɺ  ৭ʑͳύλʔϯͰΩϟονίϐʔ͕ੜ੒Ͱ͖Δ ◦ ղऍੑ΋ߴ͍: Ϟσϧ͕ͱͯ΋୯७ͳͨΊ ▪ ϩδεςΟοΫճؼͷಛ௃ྔͷॏΈΛ෼ੳ͢Ε͹ ▪ ୯ޠ-unigram: Ωϟονίϐʔʹ࢖ΘΕ΍͍͢ಛ௃తͳ୯ޠ͕Θ͔Δ ▪ ୯ޠ-ngram: จମֶ͕शͰ͖Δɻະ஌ޠॲཧΛߦ͏͜ͱͰςϯϓϨʔτ΋֫ಘͰ͖Δɻ ◦ ੜ੒͢ΔͷͰ͸ͳ͘ ධՁثΛ࡞͍ͬͯΔͷͰɺΦϖϨʔγϣϯʹରͯ͠਌࿨ੑ͕ߴ͍ ▪ Ϋϥ΢υιʔγϯάͰ͋Ε͹ɺॳֶऀͷ܇࿅ʹ࢖͑Δ ▪ ੒Ռ෺ͷϑΟϧλͱͯ͠ͷԠ༻΋ߟ͑ΒΕΔ

Slide 29

Slide 29 text

݁Ռ: Ͳ͏͍͏Ωϟονίϐʔ͕Ͱ͖Δ͔ʁ ࣾ಺ͰͷਓखධՁͰ͸ ఆྔతʹ΋ਓ͕ؒ࡞੒ͨ͠ΑΓ༗ҙʹྑ͍Ωϟονίϐʔ͕Ͱ͖Δ͜ͱ͕Θ͔ͬͨ શళฮͰ͸ແཧ͕ͩಛఆͷϑΟϧλΛ͔·ͤ͹ϓϩμΫτΠϯ΋Ͱ͖ͨ ◦ (ਓख) ౎಺࠷ڧͷ͏ͲΜ ◦ (ػց) ே͔Β൩·Ͱ௕ऄͷྻ͕Ͱ͖Δ໊ళ͏ͲΜ԰͞Μ ◦ (ػց) ೋށ࢈ͷͦ͹Λళ಺Ͱ੡ค͠ɺṢ͖ͨͯɾଧͪͨͯɾᣐͰͨͯͷʮ̏ͨͯʯͰఏڙ ◦ (ਓख) ͓ംͪΌΜͷՈʹ༡ͼʹདྷͨΑ͏ͳݹຽՈͰ௖͘ίγͷڧ͍͓ڶഴ͸ඒຯ

Slide 30

Slide 30 text

Slide 31

Slide 31 text

toCྖҬͰͷσʔλ׆༻ʢػցֶशʣͷ஌ݟ ྑ͍σʔλ͸໰୊Λγϯϓϧʹͯ͘͠ΕΔ ▪ ྑ͍໰୊ઃఆ͸ෳ਺ͷղܾΛ༩͑ͯ͘ΕΔ (Simple > Easy) ▪ ʢαʔϏεʗۀքʗλεΫʣυϝΠϯಛ༗ͷಛԽ͢Δ͜ͱͰɺΑΓ໰୊ΛγϯϓϧʹͰ͖Δ Ұఆਫ४ͷ୲อʹͱͯ΋ۤ࿑͢Δ ◦ ϞσϧʙγεςϜͷ͏·͍ύΠϓϥΠϯͱͯ͠ͷઃܭྗ͕ࢼ͞ΕΔ ◦ ΞΧσϛοΫͰ͋Ε͹ ͻͱͭͣͭͰධՁɾղܾ͢Δෳ਺ͷ໰୊Λಉ࣌ʹղܾ͢Δඞཁ͕͋Δ ◦ Ωϟονίϐʔͷ৔߹͸ ྲྀெੑ / ৴པੑʢղऍੑʣ / ॊೈੑ Λಉ࣌ʹຬͨ͢ඞཁ͕͋ͬͨ ◦ ਓͷؒҧ͍ʹ͸ൺֱతڐ༰త͕ͩɺػցతͳؒҧ͍͸ඇڐ༰త ʢਓΈ͍ͨʹؒҧ͍͑ͨʣ A/BςετͷΑ͏ͳܗͰΠϯϋ΢εͳධՁ͕ར༻Ͱ͖ΔΞυόϯςʔδ ◦ αʔϏεಛ༗ͷ݁Ռʹͳͬͯ͠·͏ͨΊɺଞͷαʔϏεʹ͓͍ͯͷ࠶ݱੑ͸୲อͰ͖ͳ͍͕…

Slide 32

Slide 32 text

toCྖҬͰͷσʔλ׆༻ʢػցֶशʣͷ೰Έ ໘ന͍ྖҬͰ΋͋Δ

Slide 33

Slide 33 text

toCͱͯ͠ޮՌతʹσʔλΛར༻͢ΔͨΊʹ͸Ͳ͏͢΂͖͔ʁ

Slide 34

Slide 34 text

͍͔ʹσʔλ׆༻Λߦ͏͔ʁ from https://simplystatistics.org/2019/04/17/tukey-design-thinking-and-better-questions/ ղ͚Δ໰୊ͷ  ೉қ౓ ղ͘΂͖໰୊ͷ  ඼࣭

Slide 35

Slide 35 text

͍͔ʹσʔλ׆༻Λߦ͏͔ʁ NNͷ୆಄ from https://simplystatistics.org/2019/04/17/tukey-design-thinking-and-better-questions/ ղ͚Δ໰୊ͷ  ೉қ౓ ղ͘΂͖໰୊ͷ  ඼࣭

Slide 36

Slide 36 text

͍͔ʹσʔλ׆༻Λߦ͏͔ʁ NNͷ୆಄ ਅͷGOAL from https://simplystatistics.org/2019/04/17/tukey-design-thinking-and-better-questions/ ղ͚Δ໰୊ͷ  ೉қ౓ ղ͘΂͖໰୊ͷ  ඼࣭

Slide 37

Slide 37 text

͍͔ʹσʔλ׆༻Λߦ͏͔ʁ NNͷ୆಄ ਅͷGOAL ཧ૝ from https://simplystatistics.org/2019/04/17/tukey-design-thinking-and-better-questions/ ղ͚Δ໰୊ͷ  ೉қ౓ ղ͘΂͖໰୊ͷ  ඼࣭

Slide 38

Slide 38 text

͍͔ʹσʔλ׆༻Λߦ͏͔ʁ NNͷ୆಄ ਅͷGOAL ཧ૝ ͜͜ʹ͍Δͭ΋Γʁ from https://simplystatistics.org/2019/04/17/tukey-design-thinking-and-better-questions/ ղ͚Δ໰୊ͷ  ೉қ౓ ղ͘΂͖໰୊ͷ  ඼࣭

Slide 39

Slide 39 text

͍͔ʹσʔλ׆༻Λߦ͏͔ʁ NNͷ୆಄ ਅͷGOAL ཧ૝ ͜͜ʹ͍Δͭ΋Γʁ ࣮ࡍ͸͔ͬͪ͜΋ʁ from https://simplystatistics.org/2019/04/17/tukey-design-thinking-and-better-questions/ ղ͚Δ໰୊ͷ  ೉қ౓ ղ͘΂͖໰୊ͷ  ඼࣭

Slide 40

Slide 40 text

͍͔ʹσʔλ׆༻Λߦ͏͔ʁ ղ͚Δ໰୊ͷ  ೉қ౓ ղ͘΂͖໰୊ͷ  ඼࣭ NNͷ୆಄ ਅͷGOAL ཧ૝ ͜͜ʹ͍Δͭ΋Γʁ ࣮ࡍ͸͔ͬͪ͜΋ʁ from https://simplystatistics.org/2019/04/17/tukey-design-thinking-and-better-questions/ ΪϟοϓΛຒΊΔ ྑ͍໰͍Λߟ͑Δඞཁ͕͋Δ

Slide 41

Slide 41 text

ྑ͍σʔλ͕͋Ε͹໰୊͸γϯϓϧʹͳΔ … ͱ͢Δͱ ྑ͍σʔλΛ͍͔ʹ࡞Δ͔Λߟ͑ΔͨΊʹ  ςΫϊϩδʔ΍ΤϯδχΞϦϯά͚ͩͰͳ͘ ྑ͍σʔλ͕ಘΒΕΔαΠΫϧΛߟ͑Δͱྑͦ͞͏

Slide 42

Slide 42 text

Krebs Cycle of Creativity https://jods.mitpress.mit.edu/pub/AgeOfEntanglement

Slide 43

Slide 43 text

ཧ૝ͷʮਓ޻஌ೳʯ͸ϧϯό…ʁ https://twitter.com/atochotto/status/1129183985119051776? ref_src=twsrc%5Etfw%7Ctwcamp%5Etweetembed&ref_url=https%3A%2F%2Fpaperusercontent.com%2Fintegrations%2Fembed%2Fiframe%2Ftweet%3Fid%3D1129183985119051776 จԽΛม͑ΒΕΔʮਓ޻஌ೳʯ͸੒ޭ ϧϯόͷத਎͕ػցֶशͷ࢓༷ͷ༗ແ͸  ۃ࿦Ͳ͏Ͱ΋͍͍͔΋͠Εͳ͍

Slide 44

Slide 44 text

Case Study: Google຋༁ ਓ޻஌ೳܥͷαʔϏε͕ʮσβΠϯʯ͔ΒʮจԽʯʹӨڹ༩͑ΔྫΛߟ࡯ͯ͠Έ͍ͨͱࢥ͏ • ̏೥͙Β͍લ͔ΒΊͪΌΑ͘ͳͬͨɻ • ͪΐ͏Ͳػց຋༁ͷύϥμΠϜ͕େ͖͘มΘΔλΠϛϯάΛ • ΞΧσϛοΫଆͰݟ͍ͯͨͷͰɺػց຋༁Ͱͷ୊ࡐΛ͋͛ͯΈ͍ͨ

Slide 45

Slide 45 text

Case Study: Google຋༁ (ௌऺͷօ͞Μʹ࣭໰) • ຋༁ීஈ࢖͍͍ͯ͠ΔਓɺखΛ͋͛ͯΈͯཉ͍͠ ◦ தֶߍ΍ߴߍͰͷ॓୊ͷࡍʹར༻͍ͯͨ͠Γ ◦ Θ͔Βͳ͍୯ޠΛࣙॻ͕ΘΓʹࡧҾ͢Δਓ ◦ Πϯλʔωοτ αʔϑΟϯͰӳޠͷχϡʔεΛ຋༁͢Δ༻్ ◦ શ͘࢖ͬͯͳ͍ਓ

Slide 46

Slide 46 text

Case Study: Google຋༁ ωλཁһʁ

Slide 47

Slide 47 text

Case Study: Google຋༁ from NLP2017νϡʔτϦΞϧʮθϩ͔Β࢝ΊΔ χϡʔϥϧωοτϫʔΫػց຋༁ʯ (http://lotus.kuee.kyoto-u.ac.jp/~nakazawa/NLP2017-NMT-Tutorial.pdf) ৽͍͠ൃ໌ or ٕज़తʹઌߦ͍ͯͨ͠ͷ͔ʁ

Slide 48

Slide 48 text

Krebs Cycle of Creativity https://jods.mitpress.mit.edu/pub/AgeOfEntanglement

Slide 49

Slide 49 text

Google຋༁͸ϓϩμΫτͱͯ͠ͳͥ੒ޭ͍ͯ͠Δ͔ʁ Α͏ͳؾ͕͢Δ

Slide 50

Slide 50 text

ڊਓͨͪ͸ͲΜͳࢹ఺Λ͍࣋ͬͯΔͷ͔

Slide 51

Slide 51 text

ڊਓͨͪͷಈ͖ʹண໨ͯ͠ΈΔ

Slide 52

Slide 52 text

ڊਓͨͪͷಈ͖ʹண໨ͯ͠ΈΔ - Netﬂix

Slide 53

Slide 53 text

ڊਓͨͪͷಈ͖ʹண໨ͯ͠ΈΔ - Netﬂix: byDevTools

Slide 54

Slide 54 text

ڊਓͨͪͷಈ͖ʹண໨ͯ͠ΈΔ - Netﬂix: ίϯςϯπ৘ใ

Slide 55

Slide 55 text

ڊਓͨͪͷಈ͖ʹண໨ͯ͠ΈΔ - Netﬂix: Πϕϯτ৘ใ - ϩέʔϧ৘ใ - ࢪࡦ൪߸৘ใ - ϦϦʔε൪߸৘ใ - Ͳͷίϯςϯπ͕ݟΒΕͨ - Ͳͷίϯςϯπ͕දࣔ͞Ε͔ͨ - ͲͷλΠϛϯάͰ  ίϯςϯπ͕ಈ͍͔ͨ …

Slide 56

Slide 56 text

ڊਓͨͪͷಈ͖ʹண໨ͯ͠ΈΔ - AirBnB

Slide 57

Slide 57 text

ڊਓͨͪͷಈ͖ʹண໨ͯ͠ΈΔ - AirBnB https://www.slideshare.net/mounialalmas/tutorial-on-online-user-engagement-metrics-and-optimization • Ϣʔβߦಈʹؔ͢ΔମܥɺͲͷΑ͏ʹଌఆ͍͔͕ͯ͘͠·ͱ·͍ͬͯΔ

Slide 58

Slide 58 text

ڊਓͨͪͷಈ͖ʹண໨ͯ͠ΈΔ - AirBnB https://www.kdd.org/kdd2018/accepted-papers/view/real-time-personalization-using-embeddings-for-search-ranking-at-airbnb ϢʔβͷৼΔ෣͍͔Βݕࡧ݁ՌʢίϯςϯπʣΛ࠷దԽ͢Δ: ηογϣϯϩάΛ Skip-gramϞσϧͰղऍͯ͠ɺίϯςϯπͱͯ͠ͷྨࣅ౓Λߏங → ݕࡧ݁Ռʹ൓ө

Slide 59

Slide 59 text

ݸਓͱͯ͠ΑΓޮՌతͳσʔλ׆༻Λ͍ͯͨ͘͠Ίʹ

Slide 60

Slide 60 text

υϝΠϯΛཧղ͢Δ͜ͱɺҙຯͷ͋ΔAIΛ࡞Δ͜ͱ͸࿈ଓత

Slide 61

Slide 61 text

Data Architectͷ͓͠͝ͱ: ྲྀ௨ͱσʔλͷܦࡁݍΛ࡞Δ͜ͱ ΞφϦετ σʔλϚʔτ ϓϩμΫτ σʔλ΢ΣΞϋ΢ε

Slide 62

Slide 62 text

RettyͰͷ׆ಈ͸͜ΕΒΛͭͳ͛Δ׆ಈ

Slide 63

Slide 63 text

ϢʔβΛཧղ͢Δ͜ͱɺ໾ʹཱͭAIΛ࡞Δ͜ͱ͸࿈ଓత σʔλ΁ͷؔΘΓํ͸΋ͬͱ৭ʑ͍͍͋ͬͯ