Temporal Search and Replace: A novel tool to simplify event sequences in large complex temporal...

16
Temporal Search and Replace: A novel tool to simplify event sequences in large complex temporal datasets Allan Fong Hanseung Lee Rongjian Lan University of Maryland Department of Computer Science

Transcript of Temporal Search and Replace: A novel tool to simplify event sequences in large complex temporal...

Page 1: Temporal Search and Replace: A novel tool to simplify event sequences in large complex temporal datasets Allan Fong Hanseung Lee Rongjian Lan University.

Temporal Search and Replace: A novel tool to simplify event sequences in large complex temporal datasets

Allan FongHanseung LeeRongjian Lan

University of MarylandDepartment of Computer

Science

Page 2: Temporal Search and Replace: A novel tool to simplify event sequences in large complex temporal datasets Allan Fong Hanseung Lee Rongjian Lan University.

Outline

Background

Our Contributions

Temporal Search and Replace Features

User Interface & Interaction Description

Demo

Search Algorithm

Conclusions and Future Work

Page 3: Temporal Search and Replace: A novel tool to simplify event sequences in large complex temporal datasets Allan Fong Hanseung Lee Rongjian Lan University.

Background

Motivation:

Simplify the visualization of large complex temporal datasets

Related works:

Search and replace research for graphics [Kurlander 1992] [Yeh 2006]

Psychology research

Miller’s 7 +/- 2 and chunking [Miller 1956]

Temporal chunking (clustering) [Farrell 2012]

Gestalt perception

(202) 456 - 2121

Page 4: Temporal Search and Replace: A novel tool to simplify event sequences in large complex temporal datasets Allan Fong Hanseung Lee Rongjian Lan University.

Our Contributions

Goal:

Develop and integrate a Temporal Search and Replace tool into EventFlow

Contributions:

Replace capabilities in EventFlow

Search capability for repeating sequences

Search capability for no ordered sequences

Search capability using wildcard notations

Customers/End Users:

Megan Monroe and Catherine Plaisant (UMD HCIL)

Sheila Weiss (UMD Baltimore)

Page 5: Temporal Search and Replace: A novel tool to simplify event sequences in large complex temporal datasets Allan Fong Hanseung Lee Rongjian Lan University.

Temporal Search and Replace Features

Type Shape & Color Visualization

Point Event Colored Triangle

Interval Event Colored Interval Shape

Any Point Event Black Triangle

Any Interval Event Black Interval Shape

Any Event Black Diamond

Repetition Black Rectangle

No Order Red Oval

≥4 { ,= ,

, ,, }

≥2= {

, …},

• One time stamp (e.g., one day)Journal Published: 5/1/10

• Two time stamps (e.g., start time and an end time)Professor: 9/5/09 - 9/30/11

Page 6: Temporal Search and Replace: A novel tool to simplify event sequences in large complex temporal datasets Allan Fong Hanseung Lee Rongjian Lan University.

Visualizing Nested Constraints

• Reverse (BFS) Breadth-First Search is used to plot inner constraint boxes & ovals first

• Dynamic add / remove constraints are easily done using the following property All nodes on the same

level are disjoint setsA parent node contains

the range of all its child nodes

Indices:

Constraints:

Events:

Page 7: Temporal Search and Replace: A novel tool to simplify event sequences in large complex temporal datasets Allan Fong Hanseung Lee Rongjian Lan University.

User Interface & Interaction

Page 8: Temporal Search and Replace: A novel tool to simplify event sequences in large complex temporal datasets Allan Fong Hanseung Lee Rongjian Lan University.

Demo

Page 9: Temporal Search and Replace: A novel tool to simplify event sequences in large complex temporal datasets Allan Fong Hanseung Lee Rongjian Lan University.

Search Algorithm

Repetition

• We extend the original EventFlow fixed-length pattern search algorithm(Temporal Pattern Search - TPS) to support flexible pattern searching.

• Search index is repeatedly reset so a pattern can be matched multiple times

No Order

• Instead of permuting the events and do sequential event matching, keep matching greedily until each event in the constraint is matched for once.

Wildcard

• Finding any event after the current time in the record

Page 10: Temporal Search and Replace: A novel tool to simplify event sequences in large complex temporal datasets Allan Fong Hanseung Lee Rongjian Lan University.

Embedded Constraints

Using tree structure to store the embedded constraints

Using stack to store the active constraints

Stack

Indices:

Constraints:

Events:

Page 11: Temporal Search and Replace: A novel tool to simplify event sequences in large complex temporal datasets Allan Fong Hanseung Lee Rongjian Lan University.

Conclusions and Future Work

Introduce a novel temporal event search and replace tool

Extend EventFlow’s search algorithm to support repetition, no order and wildcard constraints.

Conduct user study with 9 participants

Future work includes:

Add rules that better suit the purpose of simplification of temporal events.

Polish the user interface to reduce the stress of query formulation.

Pattern-level find and replace.

Page 12: Temporal Search and Replace: A novel tool to simplify event sequences in large complex temporal datasets Allan Fong Hanseung Lee Rongjian Lan University.

Acknowledgements

Megan Monroe, Catherine Plaisant, and Ben Shneiderman

Sheila Weiss

Kent Norman

And for all the participants of the usability study!

Page 13: Temporal Search and Replace: A novel tool to simplify event sequences in large complex temporal datasets Allan Fong Hanseung Lee Rongjian Lan University.

Thank you for your attention!

Question?

Page 14: Temporal Search and Replace: A novel tool to simplify event sequences in large complex temporal datasets Allan Fong Hanseung Lee Rongjian Lan University.

Backup Slides

Page 15: Temporal Search and Replace: A novel tool to simplify event sequences in large complex temporal datasets Allan Fong Hanseung Lee Rongjian Lan University.

Usability Study

9 participants (5 males, 4 females, mean age 26.7, std 2.5)

Background review, Training session, Usability testing

Search for event sequence and replace

Search for people who had both book and newspaper publications where order is not important

Search for people who have at least 10 conference publications

Search and replace all publication events

Very noticeable reduction in visual clutter after using Tool

Repetition more intuitive than No Order and Wildcard searches

Page 16: Temporal Search and Replace: A novel tool to simplify event sequences in large complex temporal datasets Allan Fong Hanseung Lee Rongjian Lan University.

Recommended Improvements

Recommendation Level of Impact

Implementation Difficulty

R1 Set a default replacement glyph for users that they can easily edit High Easy

R2 Set default interval glyphs to span the entire search event sequence High Easy

R3 Make it easier to select and drag edges of interval markers Medium Medium

R4 Reduce space between two event sequence boxes High Easy

R5 Add visualization of where replacement events are snapping to in the original sequence

Medium High

R6 Add features to let users search and incrementally replace sequences High High

R7 Combine the two-steps search and replace process into one step Medium Easy

R8 Map the flow of user clicks for all tasks to systematically eliminate unnecessary clicks

High High

R9 Activate event constraint windows with a button rather than the click and draft option or make the click and drag response less sensitive

Medium High

R10 Create a generic wildcard Medium High

R11 Include summary of the number of replacements found High Medium

R12 Add time constraints for event sequences, such as searching between two time periods

High High

R13 Make interval glyph more representative of intervals, particularly in the overview panel

Low Medium

R14 Include additional ranking features, such as ranking by interval duration or ranking by the number of events in an interval

High High

R15 Automate the alignment of event sequences after a search Low Medium