LINE TECHPULSE 2019 - Keynote

LINE TECHPULSE 2019 - Keynote

Keynote by Marco Chen @ LINE TECHPULSE 2019 https://techpulse.line.me/

2102a6b8760bd6f57f672805723dd83a?s=128

line_developers_tw

December 04, 2019
Tweet

Transcript

  1. None
  2. > > Marco Chen > > Senior Technical Director, LINE

    Taiwan Limited LINE TAIWAN TECHPULSE 2019 Keynote
  3. 2016 2019 2017 2018 10 Times Number of LINE Taiwan

    engineers has increased by 10 times since 2016 4th LINE TAIWAN TECHPULSE
  4. More Technical Insight > Almost 30 LINE Taiwan engineers shared

    technical insight in 
 DevDay 2019, Japan
  5. Life on LINE > > Connect with LINE Platform >

    > Natural Experience with AI Technology
  6. LINE SPOT Finding restaurants nearby and make reservation Life on

    LINE Featuring services for every part of your day
  7. TAIWAN JAPAN VIETNAM INDONESIA MENA Conomi ͠ͽͧ͡ NOW SHOPPING GO

    LINEϪξη GET IT LINE SHOPPING SWAY LIVE MENA
  8. Connect with LINE Platform

  9. Application publishing platform LINE MINI APP

  10. Natural Experience with AI Technology

  11. LINE is AI Company

  12. What is AI?

  13. LINE AI Beyond Clova and Chatbot

  14. AI > Machine takes over human tasks > Learn, think,

    decide, improve
  15. Fuzzy Logic, Neural Network Enble AI to Infer 
 Like

    Human Look Back on AI Development
  16. None
  17. LINE AI Application

  18. LINEਮ๐盅ݣ

  19. LINE MUSIC AI DJ

  20. SMART CHANNEL

  21. LINE TODAY > 糔֦ࡅ稭

  22. LINEਮ๐ੜ䒻ಋ

  23. LINE懱௳礚挨

  24. Training Reimburse Data Model Actual Reimburse Data Text Classifier Training

    Evaluation Data Model Creation Process Usage Process Automatically generate all documents for reimburse application LINE OCR API Mac App iOS App LINE AI Hackathon - ᛔ㵕䁭癱羬翄 ML ML
  25. 22222 LINE Taiwan Service LINE AI Data NLU Ad Targeting

    OCR Text Recognition
  26. v Next Step Face LINE AI Speech Voice

  27. Collaboration With LINE Clova AI Team LINE Clova AI team


    instructed on leveraging LINE AI technology Join Global Team Share platform and insight with global team to evolve LINE services Machine Learning System From Scratch Created local
 machine learning system to build chatbot Collaboration With LINE BRAIN Team Advance LINE AI technology with Chinese language support LINE Taiwan Data Engineering Task Force
  28. Provide for 3rd Parties Data New Services / Functions UI

    / UX Improvement Algorithms Users /Services ML Engineers Search, Recommendation, Ad Platform, Monitoring, Chatbot, Voice, Speech, Vision, OCR, Face, Video…
  29. LINE Taiwan Data Engineering Task Force and LINE Brain Advance

    LINE AI technology with Chinese language support Evolve LINE Taiwan services with global team LINE BRAIN CHATBOT LINE BRAIN TEXT ANALYTICS LINE BRAIN SPEECH TO TEXT LINE BRAIN TEXT TO SPEECH LINE BRAIN OCR LINE BRAIN VISION LINE BRAIN VIDEO ANALYTICS Text Speech Vision
  30. None
  31. Privacy First

  32. Data Governance Responsibility as a Platform Unshakable foundation for sustainable

    trustworthy of users
  33. New records 1 Trillion Hive / Spark jobs, 
 30K

    Presto queries 70K (Compressed) 
 new data volume +390 TB Every Day
  34. Implementing a scalable, on-demand analytics environment Multi-Tenancy Data Quality Automated

    data validation & meta data collection Accessibility Seamless unification of Hadoop clusters into a single big cluster Key Challenges
  35. Unified Self-Service Data Platform Data Science ML Engineering Data Governance

    Governed Self-Service Data Platform
  36. Feature as a Service Standardization Democratization Feature Engineering Y -

    features Obfuscated User Features Company-wide Components Service-specific Components Machine Learning Team Service-specific Components etc. Various Teams ML Solutions Centralized Features (Managed by ML Team) Z - features User Features C - features Content Features Feature Engineering E T L Log E T L Contents Policy Service(s)
  37. Feature Extraction List of package ids (order by timestamp) mid:

    [10, 4465, 1025960, 5413, 1456, 1299646, ...] ⁃ 10: [0.1, 0.5, 0.2, 0.2, 0.35, ...] ⁃ 4465: [0.9, 0.8, 0.4, 0.1, 0.2, ...] ⁃ 1025960: [0.45, 0.2, 0.2, 0.6, 0.9, ...] ⁃ 5413: [0.8, 0.8,0.1, 0.2, 0.5, ...] ⁃ 1456: [0.7, 0.3, 0.7, 0.3, 0.2, ...] ⁃ 10: [0.0, 0.0, 0.0, ...., 0.3, 0.1, 0.9, .., 0.0, 0.0, 0.0] ⁃ 4465: [0.3, 0.6, 0.1, …., 0.0, 0.0, 0.0, ..., 0.0, 0.0, 0.੉ ⁃ 1025960: [0.0, 0.0, 0.0, ..., 0.0, 0.0, 0.0, ..., 0.7, 0.3, 0.1] ⁃ 5413: [0.9,0.1,0.1,....,0.0, 0.0,0.0, ..., 0.0, 0.0, 0.0] ⁃ 1456: [0.0, 0.0, 0.0, ..., 0.0, 0.0, 0.0, ..., 0.2, 0.2, 0.6 mid: [0.1, 0.2, 0.1, … 0.3, 0.1, 0.8, …, 0.1, 0.6, 0.4] mid: [0.5, 0.3, 0.2, 0.9, 0.8] fastText GMM Accumulate dimensionality reduction (PCA with matrix sketching) y-sparse-features (typically 6,000 dimension) y-dense-features (typically 400 dimension)
  38. Privacy AI Data Governance

  39. Leading AI Technology