Upgrade to PRO for Only $50/Year—Limited-Time Offer! 🔥

テキスト審査 in Pairs

Avatar for goka goka
November 24, 2017

テキスト審査 in Pairs

Avatar for goka

goka

November 24, 2017
Tweet

More Decks by goka

Other Decks in Technology

Transcript

  1. ࣗݾ঺հ 2 Copyright © 2017 eureka, Inc. All rights reserved.

    CONFIDENTIAL Not for Public Distribution - Do Not Copy 2 ޙ౻ উ࠸ Katsuya Goto
  2. ϓϩμΫτ঺հ 3 Copyright © 2017 eureka, Inc. All rights reserved.

    CONFIDENTIAL Not for Public Distribution - Do Not Copy 3
  3. ձ໊ࣾ גࣜձࣾΤ΢ϨΧ / eureka, Inc. ઃཱ 2008೥11݄20೔ ୅දऀ ୅දऔక໾CEOɹੴڮ४໵ ࣾһ਺

    130໊ʢΞϧόΠτؚΉʣ ࣄۀ಺༰ ࿀Ѫɾࠗ׆ϚονϯάαʔϏεʮPairsʯͷӡӦ Χοϓϧ޲͚ίϛϡχέʔγϣϯΞϓϦʮCouplesʯͷӡӦ ؔ࿈ձࣾ IAC/InterActiveCorpʢΞϝϦΧʣ Match GroupʢΞϝϦΧʣ །ྷՈވ㟨༗ݶެ࢘ʢ୆࿷ʣ EUREKA SG Pte. Ltd.ʢγϯΨϙʔϧʣ Overview Copyright © 2017 eureka, Inc. All rights reserved. CONFIDENTIAL Not for Public Distribution - Do Not Copy 4
  4. ҆৺ ͱ ҆શ ΛकΔͨΊ Copyright © 2017 eureka, Inc. All

    rights reserved. CONFIDENTIAL Not for Public Distribution - Do Not Copy 10
  5. Copyright © 2016 eureka, Inc. All rights reserved. CONFIDENTIAL Not

    for Public Distribution - Do Not Copy 11 ۀऀͷଘࡏ ෆਖ਼Ϣʔβʔʹ͍ͭͯ Ϛονϯάͨ͠ͱࢥͬͨΒผαΠτʹ༠ಋ͞Ε·ͨ͠
  6. Copyright © 2016 eureka, Inc. All rights reserved. CONFIDENTIAL Not

    for Public Distribution - Do Not Copy 12 ۀऀͱ͸ʁ αʔϏε಺Ͱຊདྷͷ໨త͔Βҳ୤ͨ͠ ߦҝΛ͢ΔϢʔβʔ ෆਖ਼Ϣʔβʔʹ͍ͭͯ
  7. Copyright © 2016 eureka, Inc. All rights reserved. CONFIDENTIAL Not

    for Public Distribution - Do Not Copy 13 ۀऀͱ͸ʁ ͍͍Ͷ ͍͍Ͷ
  8. Copyright © 2016 eureka, Inc. All rights reserved. CONFIDENTIAL Not

    for Public Distribution - Do Not Copy 14 ۀऀͱ͸ʁ ͍͍Ͷ ͍͍Ͷ Ϛονϯάʂʂ
  9. Copyright © 2016 eureka, Inc. All rights reserved. CONFIDENTIAL Not

    for Public Distribution - Do Not Copy 15 ۀऀͱ͸ʁ ͸͡Ί·ͯ͠ʂΑΖ͓͘͠Ͷ͕͍͠·͢ʂ
  10. Copyright © 2016 eureka, Inc. All rights reserved. CONFIDENTIAL Not

    for Public Distribution - Do Not Copy 16 ۀऀͱ͸ʁ ͸͡Ί·ͯ͠ʂΑΖ͓͘͠Ͷ͕͍͠·͢ʂ ΑΖ͓͘͠Ͷ͕͍͠·͢ʂ
  11. Copyright © 2016 eureka, Inc. All rights reserved. CONFIDENTIAL Not

    for Public Distribution - Do Not Copy 17 ۀऀͱ͸ʁ ͸͡Ί·ͯ͠ʂΑΖ͓͘͠Ͷ͕͍͠·͢ʂ ΑΖ͓͘͠Ͷ͕͍͠·͢ʂ ࠷ۙ౤ࢿͷ͓ख఻͍Λ͍ͯͯ͠.. όΠφϦʔΦϓγϣϯͬͯͬ͠ ͯ·͔͢ʁࢲ͸ઌ݄30ສ΄Ͳ…
  12. Copyright © 2016 eureka, Inc. All rights reserved. CONFIDENTIAL Not

    for Public Distribution - Do Not Copy 18 ۀऀͱ͸ʁ ͸͡Ί·ͯ͠ʂΑΖ͓͘͠Ͷ͕͍͠·͢ʂ ΑΖ͓͘͠Ͷ͕͍͠·͢ʂ ͢Έ·ͤΜͪΐͬͱॾࣄ৘͋ͬ ͯɺԼه͔Β࿈བྷ͍ͩ͘͞ʂ http://hoge.com ଴ͬͯ·͢ʂ
  13. Copyright © 2016 eureka, Inc. All rights reserved. CONFIDENTIAL Not

    for Public Distribution - Do Not Copy 19 ۀऀͱ͸ʁ ͸͡Ί·ͯ͠ʂΑΖ͓͘͠Ͷ͕͍͠·͢ʂ ΑΖ͓͘͠Ͷ͕͍͠·͢ʂ ͢Έ·ͤΜͪΐͬͱॾࣄ৘͋ͬ ͯɺԼه͔Β࿈བྷ͍ͩ͘͞ʂ http://hoge.com ଴ͬͯ·͢ʂ ۀऀΛ0ʹ͢Δඞཁ͕͋Δ
  14. ҆৺ͱ҆શ 24࣌ؒମ੍ͷ؂ࢹ 365೔ͷ؂ࠪରԠCSҰ࣍ରԠ ొ ࿥ ࣌ ొ࿥৚݅ wҎԼʹ౰ͯ͸·Δͱ'BDFCPPLొ࿥ෆՄ ɹʷ'BDFCPPL༑͕ͩͪਓະຬ ɹʷطࠗɾ࿀ਓ͋Γ

    ɹʷաڈʹۀऀͱ൑ఆ͞ΕͨϢʔβʔͷ'BDFCPPL༑ͩͪ ೥ྸ֬ೝ w೥ྸ֬ೝ͕׬ྃ͠ͳ͍ͱϝοηʔδަ׵ෆՄ ɹূ໌ॻͷఏ͕ࣔඞཁ ར ༻ ࣌ ࣸਅɾ౤ߘ؂ࠪ wಠࣗج४ʹج͖ͮ࣌ؒ৹ࠪ ɹෆਖ਼Ϣʔβ௨ใɾύτϩʔϧ wෆਖ਼ͳϢʔβΛ௨ใ΍ಠࣗϩδοΫͰநग़ wਵ࣌ܯࠂୀձͷରԠΛ࣮ࢪ wར༻ऀ͔Βͷ໰͍߹Θͤʹ೔ରԠ Facebook༑ୡಉ࢜͸දࣔ͞Εͳ͍ͷͰ ҆৺ͯ͠͝ར༻͍͚ͨͩ·͢ Copyright © 2017 eureka, Inc. All rights reserved. CONFIDENTIAL Not for Public Distribution - Do Not Copy 22
  15. ҆৺ͱ҆શ ొ ࿥ ࣌ ొ࿥৚݅ wҎԼʹ౰ͯ͸·Δͱ'BDFCPPLొ࿥ෆՄ ɹʷ'BDFCPPL༑͕ͩͪਓະຬ ɹʷطࠗɾ࿀ਓ͋Γ ɹʷաڈʹۀऀͱ൑ఆ͞ΕͨϢʔβʔͷ'BDFCPPL༑ͩͪ ೥ྸ֬ೝ

    w೥ྸ֬ೝ͕׬ྃ͠ͳ͍ͱϝοηʔδަ׵ෆՄ ɹূ໌ॻͷఏ͕ࣔඞཁ Copyright © 2017 eureka, Inc. All rights reserved. CONFIDENTIAL Not for Public Distribution - Do Not Copy 23
  16. ҆৺ͱ҆શ Copyright © 2017 eureka, Inc. All rights reserved. CONFIDENTIAL

    Not for Public Distribution - Do Not Copy 24 ར ༻ ࣌ ࣸਅɾ౤ߘ؂ࠪ wಠࣗج४ʹج͖ͮ࣌ؒ৹ࠪ ɹෆਖ਼Ϣʔβ௨ใɾύτϩʔϧ wෆਖ਼ͳϢʔβΛ௨ใ΍ಠࣗϩδοΫͰநग़ wਵ࣌ܯࠂୀձͷରԠΛ࣮ࢪ wར༻ऀ͔Βͷ໰͍߹Θͤʹ೔ରԠ
  17. + γεςϜతʹରࡦ͢Δ Copyright © 2017 eureka, Inc. All rights reserved.

    CONFIDENTIAL Not for Public Distribution - Do Not Copy 27
  18. ϫʔυϑΟϧλͱ͸ Copyright © 2017 eureka, Inc. All rights reserved. CONFIDENTIAL

    Not for Public Distribution - Do Not Copy 29 ܗଶૉղੳʹΑΔ୯ޠҰக
  19. Copyright © 2017 eureka, Inc. All rights reserved. CONFIDENTIAL Not

    for Public Distribution - Do Not Copy 30 ܗଶૉղੳͷҰྫ FXͱ͔גʹڵຯ͋Γ·͔͢ʁ words: 4 FX ໊ࢺ,ݻ༗໊ࢺ,૊৫,*,*,*,* ג ໊ࢺ,Ұൠ,*,*,*,*,ג,Χϒ,Χϒ ڵຯ ໊ࢺ,Ұൠ,*,*,*,*,ڵຯ,Ωϣ΢ϛ,Ωϣʔϛ ͋Γ ಈࢺ,ཱࣗ,*,*,ޒஈɾϥߦ,࿈༻ܗ,͋Δ,ΞϦ,ΞϦ ------ non-words: 4 ͱ͔ ॿࢺ,ฒཱॿࢺ,*,*,*,*,ͱ͔,τΧ,τΧ ʹ ॿࢺ,֨ॿࢺ,Ұൠ,*,*,*,ʹ,χ,χ ·͢ ॿಈࢺ,*,*,*,ಛघɾϚε,جຊܗ,·͢,Ϛε,Ϛε ͔ ॿࢺ,෭ॿࢺʗฒཱॿࢺʗऴॿࢺ,*,*,*,*,͔,Χ,Χ ʁ ه߸,Ұൠ,*,*,*,*,ʁ,ʁ,ʁ
  20. ୯ޠҰக Copyright © 2017 eureka, Inc. All rights reserved. CONFIDENTIAL

    Not for Public Distribution - Do Not Copy 31 FXͱ͔גʹڵຯ͋Γ·͔͢ʁ FX͕BadWordͱͯ͠ొ࿥͞Ε͍ͯΕ͹ɺ ͜ͷจষ͸ෆਖ਼Β͍͠ͱ൑ఆͰ͖Δɻ
  21. Copyright © 2017 eureka, Inc. All rights reserved. CONFIDENTIAL Not

    for Public Distribution - Do Not Copy 32 ܗଶૉղੳ in Go ܗଶૉղੳϥΠϒϥϦ https://github.com/evalphobia/go-jp-text-ripper Owner͸eurekaࣾһ & αʔϏεʹ౤ೖ࣮੷͋Γ
  22. Copyright © 2017 eureka, Inc. All rights reserved. CONFIDENTIAL Not

    for Public Distribution - Do Not Copy 33 func TokenizeText(text string) tokenizer.TokenList { t := tokenizer.New() wordList, nonWordList := t.Tokenize(stripSpaceFromText(text)) validNonWords := make([]*tokenizer.Token, 0, len(nonWordList.List)) for _, w := range nonWordList.List { if utf8.RuneCountInString(w.Surface) >= 3 { validNonWords = append(validNonWords, w) } } return tokenizer.TokenList{List: append(wordList.List, validNonWords...)} } func stripSpaceFromText(text string) string { return strings.Map(func(r rune) rune { if unicode.IsSpace(r) { return -1 } return r }, text) } ܗଶૉղੳͷϙΠϯτ 3จࣈҎ্ͩͱҙຯͷ͋Δ ݴ༿ʹͳΔՄೳੑ͕͋Δ εϖʔε͸ແҙຯͳͷͰશ෦࡟আ
  23. ϫʔυϑΟϧλͱ͸ Copyright © 2017 eureka, Inc. All rights reserved. CONFIDENTIAL

    Not for Public Distribution - Do Not Copy 34 ෦෼ҰகʹΑΔείΞϦϯά
  24. ϫʔυϑΟϧλͱ͸ Copyright © 2017 eureka, Inc. All rights reserved. CONFIDENTIAL

    Not for Public Distribution - Do Not Copy 35 FX΍גʹڵຯ͋Γ·͔͢ʁ 50఺ 50఺
  25. Copyright © 2017 eureka, Inc. All rights reserved. CONFIDENTIAL Not

    for Public Distribution - Do Not Copy 36 func BadText(text string) bool { score := 0 for _, wordFilter := range wordFilters { if strings.Contains(text, wordFilter.Text) { score += wordFilter.Score } } if score >= thresholdScore { return true } return false } ϫʔυϑΟϧλͱ͸ ୯७ͳstrings.ContainsΛ࢖ͬͯείΞϦϯά
  26. ϫʔυϑΟϧλͱ͸ Copyright © 2017 eureka, Inc. All rights reserved. CONFIDENTIAL

    Not for Public Distribution - Do Not Copy 37 FͱXʹڵຯ͋Γ·͔͢ʁ 100఺ Θ͟ͱ୯ޠʹͤͣɺܗଶૉղੳରࡦ͞Εͨηϯςϯε͕ ଟ͍ͷͰɺͪ͜Β΋༗ޮɻ
  27. ϫʔυϑΟϧλͱ͸ Copyright © 2017 eureka, Inc. All rights reserved. CONFIDENTIAL

    Not for Public Distribution - Do Not Copy 38 ੑผຖʹϫʔυΛొ࿥Ͱ͖ΔΑ͏ʹ͠ɺಛ௃తͳϫʔυʹ΋ରԠ͍ͯ͠Δɻ ྫʣԶͷ໊લ͸FX => ͜Ε͸உੑϢʔβʔ༻ʹ͔͠ొ࿥͠ͳ͍ͳͲɻ ո͍͠ςΩετͷID͸ΧελϚʔέΞνʔϜʹSlackܦ༝Ͱ௨ใ͞Εɺ͙͢ʹ ৹ࠪ݁ՌΛೖྗͯ͠΋Β͏ɻશମͷޡΓ཰Λߟྀͯ͠ɺϫʔυͷείΞΛԼ ͍͛ͯΔɻ
  28. ୯ޠҰக Copyright © 2017 eureka, Inc. All rights reserved. CONFIDENTIAL

    Not for Public Distribution - Do Not Copy 39 ࠓ೔঺հͨ͠ϫʔυϑΟϧλҎ֎ʹ΋ ϫʔυͷϗϫΠτϦετ࡞੒ ΰϦΰϦͷਖ਼نදݱ ػցֶश etc… ͰϢʔβʔͷ҆શΛक͍ͬͯ·͢ʂ
  29. Making a great impact on people’s life. * Gopher designed

    by Hinako Sakawa - Original Gopher designed by Renee French. https://github.com/ohina