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
   
   

   View Slide

2. Outline
   •  Motivation & Challenges
   •  Problem Formulation
   •  Our Solution
   •  Evaluation & Insights
   

   View Slide

3. Motivation: Query Specification
   Specification Execution
   Specification
   Amount of Time
   Data Processing Tasks
   Data Analysis Tasks
   Execution
   

   View Slide

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
   

   View Slide

5. Motivation:"
   “Next-generation” Interfaces
   •  Touch, Gestures, and Motion-capture
   •  iPad, Kinect, HoloLens, Leap Motion, Surface
   Hub, BMW Cars
   •  Direct manipulation
   •  Better interactivity
   

   View Slide

6. Humans are bad at querying!
   

   View Slide

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!
   ✔!
   ✗!
   

   View Slide

8. How do we guide the user to the
   “right” query by providing
   interactive feedback?
   ANSWER:
   "
   Snap To Queries during
   Exploratory Specification Step
   

   View Slide

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
   

   View Slide

10. Outline
    •  Motivation & Challenges
    •  Problem Formulation
    •  Our Solution
    •  Evaluation & Insights
    

    View Slide

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
    

    View Slide

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
    

    View Slide

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
    

    View Slide

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)>∆'!?!
    !?!
    

    View Slide

15. Snapping Function
    •  Linear snapping function
    •  Other snapping functions (sigmoid, exponential, etc)
    (="​∆"−"∆'/∆ℎ"−"∆' +"∆'
    !"=("×'ℎ
    
    

    View Slide

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
    

    View Slide

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
    

    View Slide

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
    

    View Slide

19. Outline
    •  Motivation
    •  Problem Formulation
    •  Our Solution
    •  Snapping Feedback
    •  Data Reduction
    •  SnapToQuery in Action
    •  Evaluation & Insights
    

    View Slide

20. SnapToQuery in Action
    

    View Slide

21. Outline
    •  Motivation
    •  Problem Formulation
    •  Our Solution
    •  Snapping Feedback
    •  Data Reduction
    •  SnapToQuery in Action
    •  Evaluation & Insights
    

    View Slide

22. Evaluation: Questions
    •  Performance
    •  Are we fast enough to provide interactive
    speeds?
    
    •  Effectiveness
    •  Does SnapToQuery help end-users?
    
    

    View Slide

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!
    

    View Slide

24. does snapping really help "
    query specification for end-users?
    

    View Slide

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
    

    View Slide

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.
    

    View Slide

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 !
    

    View Slide

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
    

    View Slide

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
    

    View Slide

30. Thank you!
    interactive data systems
    research group at the ohio state university
    papers, videos & more at http://interact.osu.edu
    

    View Slide