Big Data, We Have A Communication Problem

This 2013 TTI/Vanguard Conference on Ginormous Systems presentation explores the cognitive bottleneck we face in working with big data -- the challenge of designing technical solutions that effectively leverage human cognitive ability.

Transcript

1. BIG DATA, We have a communication problem. GINORMOUS SYSTEMS April 30–May 1, 2013 Washington, D.C. Daniel Tunkelang Head of Query Understanding, LinkedIn
2. BIG DATA IS EVERYWHERE
3. BIG DATA POWERS EVERYTHING
4. DATA SCIENTISTS WORRY ABOUT VOLUME, VELOCITY, VARIETY, …
5. BUT THE BOTTLENECK ISN’T COMPUTATIONAL IT’S COGNITIVE
6. TOOLS AUGMENT HUMAN INTELLECT BIG DATA IS A TOOL Doug Engelbart, inventor of the mouse, hypertext, etc.
7. NOT EVERYONE SUBSCRIBES TO THIS POINT OF VIEW… Claudia Perlich, Chief Scientist of media6degrees, speaking at TTI/Vanguard 2012 Conference on Understanding Understanding:
8. SHE HAS A POINT
9. BUT PREDICTIVE MODELING IS NOT ENOUGH
10. TRAINING DATA? OBJECTIVE FUNCTION?
11. WE NEED A PEOPLE-CENTRIC APPROACH TO BIG DATA INTERPRETABILITY INTERACTION INSIGHT
12. LET’S START WITH INTERPRETABILITY
13. EXAMPLE: SVM vs. DECISION TREE
14. DECISION TREES HAVE FLAWS… DISCRET E
15. BUT THEY COMMUNICATE (if they’re shallow) early splits provide big picture… fat leaves guide feature engineering …or reveal training data problems
16. WHICH SUPPORTS ITERATION
17. INTERPRETABILITY DELIVERS Key search leader favors rule-based approach for key scoring algorithms. Replaced regression with decision tree in local search model: gained accuracy and insight. Using trees to recognize spam, analyze search abandonment, model / quantify social proof.
18. GO DEEP vs INTERPRETABILITY A KEY DATA SCIENCE TRADE-OFF
19. ON TO INTERACTION
21. DON’T OVERPAY FOR PRECISION
22. BE FAST, CHEAP, AND 98% RIGHT http://metamarkets.com/2012/fast-cheap-and-98-right-cardinality-estimation-for-big-data/
23. ARE PEOPLE THAT IMPATIENT? tolerable wait time for web users 0.1s increase in latency significantly reduces # of searches, ad revenue tl;dr: YES
24. IMPATIENCE IS GOOD SPEED MATTERS
25. INSIGHT
26. http://blog.takejune.com/archives/52334044.html
27. BE TRENDY AND NORMALIZE vs
28. Sept. 11th Abu Ghraib Weapons Inspectors SOLVE FOR INTERESTINGNESS
29. COMPUTE POTENTIAL INSIGHTS APPLY HUMAN INTUITION
30. SUMMARY: Let’s have a conversation with Big Data. INTERPRETABILITY INTERACTION INSIGHT