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

COVID-19 に対するオープンデータの取り組み

COVID-19 に対するオープンデータの取り組み

第44回 Machine Learning 15minutes! で登壇した際の資料です

Asei Sugiyama

April 25, 2020
Tweet

More Decks by Asei Sugiyama

Other Decks in Technology

Transcript

  1. COVID-19 ʹؔ͢Δ
    ΦʔϓϯσʔλͷऔΓ૊Έ
    ୈ 44 ճ Machine Learning 15minutes!
    1

    View Slide

  2. ࣗݾ঺հ
    • ਿࢁ Ѩ੟
    • Software Engineer @ Repro AI Labs
    • TensorFlow ίϯτϦϏϡʔλʔ
    • TFX : Issue ཱͯͨΓ PR ཱͯͨΓ
    • docs-l10n : ຋༁ & ϨϏϡʔ
    • ػցֶशਤؑ ڞஶ
    2

    View Slide

  3. ஫ҙࣄ߲
    • ͋͘·ͰݸਓతͳݟղͰ͋Γɺॴଐ͢Δ͍͔ͳΔஂମͱ΋ແؔ
    ܎Ͱ͢
    • σʔλαΠΤϯεͷ࿩͸͋·Γ͠·ͤΜ
    • ػցֶशͷ࿩͸͠·ͤΜ
    • COVID-19 ʹؔ͢Δকདྷ༧ଌͷ࿩͸͠·ͤΜ
    • COVID-19 ͷײછ༧๷ͷ࿩͸͠·ͤΜ
    3

    View Slide

  4. ຊ೔ͷ಺༰
    1.Open Data ͱ͸ <-
    2.COVID-19 ʹؔ͢Δ Open Data ͱͦͷ׆༻
    3.զʑ͸σʔλͰԿ͕Ͱ͖Δͷ͔
    4

    View Slide

  5. Open Data ͱ͸
    • “Open data and content can be
    freely used, modified, and
    shared by anyone for any
    purpose” from The Open
    Definition
    • ϚγϯϦʔμϒϧͰ͋Δ͜ͱ΋ॏཁࢹ
    • Data.gov : ΞϝϦΧͷΦʔϓϯσʔ
    λͷαΠτ (ӈਤ)
    • Data Portals : 588 ͷΦʔϓϯ
    σʔλͷΧλϩά
    5

    View Slide

  6. ೔ຊͷ Open Data
    • ࣏ࣗମ͕ؤு͍ͬͯΔ
    • data.go.jp
    • σʔλγςΟḉߐϙʔλϧαΠτ
    • ͍͢͝ (ޙड़)
    • ౦ژ౎ΦʔϓϯσʔλΧλϩάαΠτ
    ϗʔϜϖʔδ
    6

    View Slide

  7. ΦʔϓϯσʔλΛ͸͡ΊΑ͏1
    • ੓෎΍ಠཱߦ੓๏ਓɺ࣏ࣗମͳͲ͕อ
    ༗͢Δެڞσʔλ͕ɺࠃຽ΍اۀͳͲ
    ͷར׆༻͞Ε΍͍͢Α͏ʹػձ൑ಡʹ
    దͨ͠ܗͰɺೋ࣍ར༻Մೳͳϧʔϧͷ
    ΋ͱͰެ։͞ΕΔ͜ͱɺ·ͨɺͦͷΑ
    ͏ʹެ։͞ΕͨσʔλΛࢦ͠·͢
    1 http://www.kantei.go.jp/jp/singi/it2/densi/kettei/
    opendata_tebikisyo.pdf
    7

    View Slide

  8. σʔλγςΟḉߐϙʔ
    λϧαΠτ
    • ͞·͟·ͳσʔλΛ PDF Ͱ͸ͳ͘
    CSV, XML, RDF, Linked RDF Ͱެ
    ։͍ͯ͠Δ
    • ϥΠηϯε͸ CC BY
    • ྉཧϨγϐ (XML)
    8

    View Slide

  9. 9

    View Slide

  10. ͞͹͑஍৔࢈໺ࡊͷྉཧϨγϐ (ൈਮ)
    ݟ΍͢͞ͷͨΊɺ࠷ޙͷߦʹ͚ͩվߦΛՃ͑ͨ

    "2"@ja ;
    "΄͏ΕΜ૲ͱ΋΍͠ͷφϜϧ"@ja ;
    "4ਓ෼"@ja ;
    "΋΍͠ 170gʢ2/3ାʣ"@ja ;
    "΄͏ΕΜ૲ 130gʢ2/3ଋʣ"@ja ;
    "͕ܲΒεʔϓͷૉ খ͞͡2"@ja ;
    "͝·༉ খ̍͞͡"@ja ;

    .
    10

    View Slide

  11. ͞͹͑஍৔࢈໺ࡊͷྉཧϨγϐ
    • http://linkdata.org/work/
    rdf1s8262i
    • ߹ܭμ΢ϯϩʔυ਺ 4 ճ
    (2020-04-23 ࣌఺)
    • ࠷ऴߋ৽೔ : 2020/03/17
    • Text, Excel, RDF ܗࣜ
    11

    View Slide

  12. ຊ೔ͷ಺༰
    1.Open Data ͱ͸
    2.COVID-19 ʹؔ͢Δ Open Data ͱͦͷ׆༻ <-
    3.զʑ͸σʔλͰԿ͕Ͱ͖Δͷ͔
    12

    View Slide

  13. COVID19 ʹؔ͢Δ Open Data ͱͦͷ׆༻
    ٸ଎ʹσʔλެ։ɾ׆༻ͷಈ͖͕޿͕͍ͬͯΔɻ
    • ౦ژ౎ ৽ܕίϩφ΢Πϧεײછ঱ରࡦαΠτ
    • ৽ܕίϩφ΢ΠϧεରࡦμογϡϘʔυ
    13

    View Slide

  14. ౦ژ౎ ৽ܕίϩφ΢
    Πϧεײછ঱ରࡦαΠ
    τ
    • ౦ژ౎ΦʔϓϯσʔλΧλϩάαΠτͷ
    σʔλ͔Β࡞੒
    • αΠτࣗମ΋ MIT ϥΠηϯε
    • Code for Japan ͷ׆ಈͷ 1 ͭ
    14

    View Slide

  15. 15

    View Slide

  16. StopCovid19 શࠃ൛
    • ౦ژ౎ͷΦʔϓϯιʔεΛར༻ͨ͠αΠ
    τͷ·ͱΊαΠτ
    • ͜ͷαΠτࣗମ΋ MIT ϥΠηϯε
    • codeforjapan/stopcovid19
    16

    View Slide

  17. ৽ܕίϩφ΢Πϧεର
    ࡦμογϡϘʔυ
    • ෱໺ହհ (@taisukef) ࢯ
    • COVID-19 ͷױऀ͕ར༻Մೳͳපচ਺
    ʹண໨ͨ͠αΠτ
    • ްੜ࿑ಇলఏڙͷ৽ܕίϩφ΢Πϧε
    ཅੑױऀ਺ PDF σʔλΛ JSON-API
    Խͨ݁͠ՌΛར༻
    • ιʔε͸CC BYϥΠηϯεͰఏڙ͞Εͯ
    ͍Δ
    17

    View Slide

  18. ౎ಓ෎ݝผ৽ܕίϩφ΢Πϧεײછऀ਺Ϛοϓ
    • δϟοάδϟύϯגࣜձࣾ
    • σʔλ͸ CC-BY-NC 4.0
    • ༷ʑͳσʔλιʔε͔Βूܭ͍ͯ͠Δ໛
    ༷ (ෳ਺ਓʹΑΔ࡞ۀ)
    • ՄࢹԽ͸ ArcGIS (ESRI)
    18

    View Slide

  19. தͷਓͷ੠
    • https://twitter.com/
    oohamazaki/status/
    1253306592595832835?s=20
    19

    View Slide

  20. SIGNATE COVID-19
    νϟϨϯδ
    • ϑΣʔζ 1. σʔληοτͷߏங
    • ϑΣʔζ 2. σʔλ෼ੳ
    • ϑΣʔζ 3. ጶױऀ਺༧ଌ
    20

    View Slide

  21. 21

    View Slide

  22. ຊ೔ͷ಺༰
    1.Open Data ͱ͸
    2.COVID-19 ʹؔ͢Δ Open Data ͱͦͷ׆༻
    3.զʑ͸σʔλͰԿ͕Ͱ͖Δͷ͔ <-
    22

    View Slide

  23. ཁࢫ
    • σʔλ͕࣍ʑʹެ։͞Ε͍ͯΔ͕ɺඇઐ໳ՈͰ͋ΔҰൠࢢຽ͕
    ݟͯಘΒΕΔ΋ͷ͸গͳ͍
    • ૬ஊ݅਺ͷരൃతͳ૿େ͕ͳ͍ͱ͍͏͜ͱ͸ɺ૭ޱ͕ർฐͯ͠
    ͍ΔՄೳੑΛࣔ͢
    • Ұൠࢢຽͱͯ࣍͠ʹ Q&A Λ੔උ͠ɺͦΕΛΞΫηε͠΍͍͢
    ܗͰఏڙͰ͖ΔΑ͏ʹ͢Δ͜ͱΛఏҊ͍ͨ͠
    23

    View Slide

  24. ݸਓతͳ҆৺
    • COVID-19 ʹؔ͢Δ AI ສೳ࿦͕ग़ͯ
    ͍ͳ͍͜ͱ
    • ༧ଌϞσϧʹΑΔ༬ݴऀ͕࣍ʑʹݱΕ
    Δঢ়گͰ͸ͳ͍ͷ͸ෆ޾தͷ޾͍
    • σʔληοτͷ੔උͱ͍͏΋ͬͱ΋ॏ
    ཁͳՕॴʹϦιʔεׂ͕͔Ε͍ͯΔ
    24

    View Slide

  25. ػցֶश͸໾ʹཱͭͷ͔ʁ
    • ػցֶश·Ͱͷεςοϓʹ͓͍ͯɺݱࡏ͸ۀ຿ͷఆٛͷஈ֊
    1.ۀ຿͕ະఆٛ
    2.ۀ຿͕ఆٛ͞ΕɺखॱॻΛ࡞Δ
    3.खॱ͕໌֬ʹͳΓɺγεςϜԽΛߦ͏ <- ࠓ͸͜ͷ͋ͨΓ
    4.৘ใΛՄࢹԽ͠ɺҙࢥܾఆΛߦ͏ <- ઐ໳Ո͸͜͜
    5.ҙࢥܾఆͷࣗಈԽ or ิॿ <- ػցֶश͸͜͜
    25

    View Slide

  26. σʔλαΠΤϯςΟετ͸໾ʹཱͭͷ͔ʁ
    • σʔλͷ׆༻ʹඞཁͳεΩϧͷ෼໺
    • σʔλαΠΤϯε
    • ΤϯδχΞϦϯά
    • υϝΠϯ஌ࣝ
    • ҰൠͷσʔλαΠΤϯςΟετɾΤϯδχΞ͸ҩֶ΍ެऺӴੜ
    ͷ஌͕ࣝඞͣ͠΋๛෋ͱ͍͏Θ͚Ͱ͸ͳ͍
    26

    View Slide

  27. ࣗ෼ʹ͸Կ͕Ͱ͖ΔͷͩΖ͏͔ʁ
    • COVID-19 ͷײછͷ֦େঢ়گʹ͍ͭͯूΊΒΕͨσʔλ͸ɺඇ
    ઐ໳Ոͷखʹ༨Δ
    • Titanic ͷσʔλ͔Βɺզʑ͸Կ͔༗༻ͳ஌ݟΛಘΒΕͨͩ
    Ζ͏͔ʁ
    • υϝΠϯ஌ࣝΛൈ͖ʹɺ਺ཧϞσϧ΍ػցֶशϞσϧΛ༻͍ͯ
    কདྷ༧ଌΛͨ͠ͱ͜ΖͰɺͦͷ݁Ռ͸Կ͔ҙຯ͕͋ΔͩΖ͏
    ͔ʁ
    27

    View Slide

  28. Amazon (AWS) ͷύ
    ϒϦοΫσʔλϨΠΫ
    • σʔλʹ͸ΞΫηε͠΍͘͢ͳͬͨ
    • ϚΫϩͳҙࢥܾఆ͸͔֬ʹͰ͖Δ͔΋
    ͠Εͳ͍
    • զʑ͸͜ΕΛݟͯԿΛ͢Ε͹͍͍Μͩ
    Ζ͏͔ʁ
    28

    View Slide

  29. See how your community
    is moving around
    differently due to
    COVID-192
    • Google ͷϨϙʔτ
    • Transit stations ΍ Workplace
    ΛݮΒͤ͹ྑͦ͞͏ͳײ͕͢͡Δ
    • ײછऀ਺ΑΓ͸མͪண͍ͯݟΒΕΔ
    2 Google LLC "Google COVID-19 Community Mobility Reports."
    https://www.google.com/covid19/mobility/ Accessed:
    2020-04-24.
    29

    View Slide

  30. զʑ͕ݟΔ΂͖ࢦඪ͸
    ͳʹ͔ʁ
    • ࣍ʹ΍Δ͜ͱ͕ݟ͑ΔࢦඪΛݟ͍ͨ
    • ͜Ε͸Ͳͷۀ຿ʹண໨͢΂͖͔ͱ͍͏
    ໰ʹಉ͡
    • ໰͍߹Θͤۀ຿ʹண໨͢Δ͜ͱΛఏҊ
    ͍ͨ͠
    • ౦ژ౎ ৽ܕίϩφίʔϧηϯλʔ૬ஊ
    ݅਺ͷྦྷܭ͕ઢܗʹ͔͠৳ͼ͍ͯͳ͍
    30

    View Slide

  31. 31

    View Slide

  32. ࣍ʹ੔උ͢΂͖σʔλ͸ͳʹ͔ʁ
    • ඞཁͳͷ͸໰͍߹Θͤͱͦͷճ౴
    • ༷ʑͳରࡦ͕ଧͨΕ͍ͯΔҰํɺ૭ޱ΁ͷΞΫηε͸໌֬Ͱͳ
    ͍ (෺ཧతʹ໾ॴʹߦ͚ͳ͍ʂ) ͷͰ͜͏͍ͬͨ૭ޱʹؔ܎ͳ
    ͍໰͍߹Θ͕ͤࡴ౸͍ͯ͠Δ͜ͱ͕༧૝͞ΕΔ
    • Q&A σʔλΛ༻͍ͨݕࡧΤϯδϯ͸໌Β͔ʹ༗ޮ
    • Q&A σʔλͱγφϦΦΛݩʹͨ͠νϟοτϘοτ͸༗ޮ
    32

    View Slide

  33. 33

    View Slide

  34. https://corona.go.jp/ ʹઃஔࡁ
    34

    View Slide

  35. ૝ఆ͞ΕΔ՝୊
    • ηϯγςΟϒͳ࣭໰͕૬౰ྔدͤΒΕ
    Δ͜ͱ͕૝ఆ͞Εɺੜσʔλͷอ؅͸
    Ͱ͖ͳ͍
    • ঢ়گ͕໨·͙Δ͘͠มԽ͢Δ݁Ռɺಉ
    ࣭͡໰ʹରͯ͠΋ద੾ͳճ౴͕ৗʹม
    ΘΓଓ͚ΔͨΊɺσʔληοτͷߋ৽
    ͕ඞཁ
    • લऀ͸ͱ΋͔͘ޙऀʹ͍ͭͯ͸ࢢຽͰ
    Ͱ͖Δ͜ͱ͕͋ΔͷͰ͸
    35

    View Slide

  36. ͳͥզʑ͸ϚεΫΛٻ
    Ίͯ͠·͏ͷ͔
    • ڴҖʹରͯ͠Ͱ͖Δ͜ͱ͕ͳʹ΋ͳ͍
    ͱ͍͏ετϨε͔ΒճආͰ͖Δ
    • ϚεΫ͸ങ͍ٻΊΔ͜ͱ͕Ͱ͖Δͷ
    Ͱɺڻҟʹରͯ͠औΕΔΞΫγϣϯʹ
    ͳ͍ͬͯΔ
    http://www.osaka-kyoiku.ac.jp/
    ~toyosawa/klecture09.pdf
    36

    View Slide

  37. σʔληοτͷ੔උ͸ࢢຽ͕ߦ͑ΔΞΫγϣϯ
    • ඇઐ໳ՈͰ΋ࢀՃͰ͖Δ
    • ࡞Δ͜ͱΑΓ΋อक͢Δ͜ͱʹଟେͳख͕͔͔ؒΔ (ϦϯΫ੾
    Ε΍σʔλͷܗࣜมߋ΁ͷରॲͱ͔)
    • σʔληοτ͕͋Ε͹ΤϯδχΞͷྗͰιϑτ΢ΣΞʹͰ͖Δ
    • ιϑτ΢ΣΞʹͰ͖Ε͹֦ࢄͰ͖Δ
    • SIGNATE ͷऔΓ૊Έ͸͋Δҙຯཧ૝త
    37

    View Slide

  38. Recap
    • σʔλ͕࣍ʑʹެ։͞Ε͍ͯΔ͕ɺඇઐ໳ՈͰ͋ΔҰൠࢢຽ͕
    ݟͯಘΒΕΔ΋ͷ͸গͳ͍
    • ૬ஊ݅਺ͷരൃతͳ૿େ͕ͳ͍ͱ͍͏͜ͱ͸ɺ૭ޱ͕ർฐͯ͠
    ͍ΔՄೳੑΛࣔ͢
    • Ұൠࢢຽͱͯ࣍͠ʹ Q&A Λ੔උ͠ɺͦΕΛΞΫηε͠΍͍͢
    ܗͰఏڙͰ͖ΔΑ͏ʹ͢Δ͜ͱΛఏҊ͍ͨ͠
    38

    View Slide