SnapToQuery: Providing Interactive Feedback during Exploratory Query Specification

Slides presented at VLDB 2015, full paper PDF available at: http://www.vldb.org/pvldb/vol8/p1250-jiang.pdf

Video of SnapToQuery in action: http://go.osu.edu/snaptoquery

A critical challenge in the data exploration process is discovering and issuing the “right” query, especially when the space of possible queries is large. This problem of exploratory query specification is exacerbated by the use of interactive user interfaces driven by mouse, touch, or next-generation, three-dimensional, motion capture-based devices; which, are often imprecise due to jitter and sensitivity issues. In this paper, we propose SnapToQuery, a novel technique that guides users through the query space by providing interactive feedback during the query specification process by “snapping” to the user’s likely intended queries. These intended queries can be derived from prior query logs, or from the data itself, using methods described in this paper. In order to provide interactive response times over large datasets, we propose two data reduction techniques when snapping to these queries. Performance experiments demonstrate that our algorithms help maintain an interactive experience while allowing for accurate guidance. User studies over three kinds of devices (mouse, touch, and motion capture) show that SnapToQuery can help users specify queries quicker and more accurately; resulting in a query specification time speedup of 1.4× for mouse and touch-based devices and 2.2× for motion capture-based devices.

Transcript

1. SnapToQuery: Providing Interactive Feedback during Exploratory Query Specification Lilong Jiang, Arnab Nandi
 Computer Science & Engineering The Ohio State University interactive data systems research group at ohio state
2. Outline •  Motivation & Challenges •  Problem Formulation •  Our Solution •  Evaluation & Insights
3. Motivation: Query Specification Specification Execution Specification Amount of Time Data Processing Tasks Data Analysis Tasks Execution
4. Motivation: " Exploratory Query Specification •  Highly Interactive Applications •  Web, Mobile •  Scientific, Business •  E.g., Parameter Exploration 
 and Tuning •  Time to specify a query is often more than time to execute query Parameter Panel
5. Motivation:" “Next-generation” Interfaces •  Touch, Gestures, and Motion-capture •  iPad, Kinect, HoloLens, Leap Motion, Surface Hub, BMW Cars •  Direct manipulation •  Better interactivity
6. Humans are bad at querying!
7. Challenges: Interactive UIs •  Lack of Precision •  Sensitivity / Jitter •  Exploration 
 is Result Agnostic 0" 0.1" 0.2" 0.3" 0.4" 0.5" 0.6" 0.7" 0.8" 0.9" 1" 0" 5000" 10000" 15000" 20000" 25000" Position! Time (ms)! Y! X! ✔! ✗!
8. How do we guide the user to the “right” query by providing interactive feedback? ANSWER: " Snap To Queries during Exploratory Specification Step
9. “Snap” Effect •  Alter relative position between cursor and handle to simulate friction and attraction •  Encourage selection of certain values •  But not disallow selection of others Snap Points
10. Outline •  Motivation & Challenges •  Problem Formulation •  Our Solution •  Evaluation & Insights
11. SnapToQuery Problem Find a representative query Qi such that 
 Result(Qi±δ ) yields same / similar result as 
 Result(Qi ) •  “Snap” to Qi •  Feedback during user interaction
12. SELECT COUNT(*) from T
 WHERE v11 < d1 < v12 , v21 < d2 < v22 , …, vj1 < dj < vj2 SNAPTO Δ ≥ Δl FEEDBACK snapping(Δ, Δl , Δh ) Query class: " n-dim range selection
13. SnapToQuery Architecture 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 DB Mouse, Tablet, Leap Motion Naive Method Data Contour Method Data Reduction Visualization Snapping Algorithm Network
14. Feedback: SnapToQuery •  Bias representative queries during interaction •  Representative queries: value is aligned with the bin boundary •  Alter relative position between cursor and handle to simulate friction and attraction gap area catch up snapping !" qr1 qr2 ∆"($%1,$%2)>∆'!?! !?!
15. Snapping Function •  Linear snapping function •  Other snapping functions (sigmoid, exponential, etc) (="​∆"−"∆'/∆ℎ"−"∆' +"∆' !"=("×'ℎ
16. Scaling Challenge •  Exploratory Specification = Frontend •  It is hard to keep all data (to compute result) at the frontend! •  Solution: Data Reduction •  Frontend uses 
 a summarized 
 representation 1 3 5 7 9 11 13 15 DB Mouse, Tablet, Leap Motion Naive Method Data Contour Method Data Reduction Visualization Snapping Algorithm Network
17. Data Reduction Strategies •  Fast to compute •  Performed offline, at backend •  Lightweight •  Used at frontend •  Fluid Experience •  Accurate •  Correctly guide the user •  Approaches Considered (details in the paper): •  Bin / Grid-based •  Sampling •  Data Contour
18. Data Contour: Snap to cluster boundary -74.02 -74.01 -74 -73.99 -73.98 -73.97 -73.96 -73.95 -73.94 start station longitude q1i q1r q1i is represented by q1r v1 q2r Snapping will happen if Δ(q1r , q2r ) > Δl v2 v
19. Outline •  Motivation •  Problem Formulation •  Our Solution •  Snapping Feedback •  Data Reduction •  SnapToQuery in Action •  Evaluation & Insights
20. SnapToQuery in Action
21. Outline •  Motivation •  Problem Formulation •  Our Solution •  Snapping Feedback •  Data Reduction •  SnapToQuery in Action •  Evaluation & Insights
22. Evaluation: Questions •  Performance •  Are we fast enough to provide interactive speeds? •  Effectiveness •  Does SnapToQuery help end-users?
23. Performance: Time-to-task 1 10 100 1000 10000 100000 logTime(µs) Original Naive Sample Data Contour DATAROAD w/ 20 bins DATAROAD w/ 50 bins DATABIKE w/ 50 bins DATABIKE w/ 20 bins •  The data reduction can help maintain an interactive performance. •  The data contour achieves the lowest time!
24. does snapping really help " query specification for end-users?
25. User Study Setup •  Interfaces: Mouse, Touch, Leap Motion •  Users: 30 users, between-subjects study •  Task o  Range Selection using Sliders o  Each task includes 3 subtasks on three sliders o  The first target is the left snapping position o  The second target is the right snapping position o  The third target is the second left snapping position •  Compare the case w/ and w/o snapping o  specification time o  miss time: how many times users miss the target query
26. Query Specification Experiments 0 20 40 60 80 100 120 140 160 mouse w/o snapping mouse w/ snapping iPad w/o snapping iPad w/ snapping leap motion w/o snapping leap motion w/ snapping Time (s) switchtime subtime3 subtime2 subtime1 1.4 X 1.4 X 2.2 X •  1.4 x speedup for the mouse and touch, 2.2 x speedup for the leap motion. •  The performance of leap motion w/ snapping is comparable to the mouse and iPad.
27. Query Specification: Accuracy" time wasted in missing the target 0 5 10 15 20 25 mouse w/o snapping mouse w/ snapping iPad w/o snapping iPad w/ snapping leap motion w/o snapping leap motion w/ snapping Miss Times submisstimes3 submisstimes2 submisstimes1 Miss times are reduced !
28. Outline •  Motivation •  Problem Formulation •  Our Solution •  Evaluation & Insights •  Data Contour method •  achieves interactive latencies •  SnapToQuery allows users to specify queries •  faster •  with better accuracy
29. Summary •  Exploratory Query Specification 
 is popular but challenging •  important to guide users •  Interactive Feedback is very useful •  take advantage of data •  data reduction using contour method •  generalizable problem formulation •  many open problems! •  SnapToQuery 
 Speeds up Query Specification Times •  1.4 x for mouse and touch, 2.2 x for leap motion
30. Thank you! interactive data systems research group at the ohio state university papers, videos & more at http://interact.osu.edu