University of Maryland

1
University of Maryland

• 1) Sonification of Maps
• 2) Viewing Search Results with Stable
  Meaningful Hierarchies
• 3) Narrated Demos
• 4) Visualization for re-identifiability
• Threads Visualization Universal
  Usability

2
1) Sonification of Maps
One Challenge use sound to present geographical
data distribution patterns to visually impaired
users Universal usability avoid special devices
3
1) Sonification of Maps

• Goals use non-speech audio to convey
• patterns in choropleth maps
• Initial prototype
• Two user studies (9 40 sighted users)
• Work with 2 blind users
• PhD work of Haixia Zhao

• Collaborators
• Ben Smith Kent Norman (Experiments, UMD)
• Ramani Duraiswami Dimitry Zotkin (Spatial sound
  production, UMD)

4
1) Sonification of Maps

• Goals use non-speech audio to convey
• patterns in choropleth maps
• Initial prototype
• Two user studies (9 40 sighted users)
• Work with 2 blind users
• PhD work of Haixia Zhao

• Zhao, et al. (2005), Interactive Sonification
  of Choropleth Maps Design and Evaluation, to
  appear in IEEE Multimedia Special Issue on
  Interactive Sonification , Apr-Jun 2005
• Zhao et al. (2004), Sonification of
  geo-referenced data for auditory information
  seeking design principle and pilot study , in
  Proc. 10th Intl Conference on Auditory Display,
  Sydney, Australia
• Zhao et al. (2003), Improving accessibility and
  usability of geo-referenced statistical data,
  Proc. Digital Government Research Conference,
  March 2003, 147-155

5
Controlled Experiments

• Subjects listened to and explored sonified data
  and select pattern among choices
• Compared map vs. table
• Compared 2 navigation methodsand 2 sound designs

6
Two Controlled Experiments

• Tasks listened to and explored sonified data and
  chose matching patterns from visual choices
• Compared map vs. table
• Compared 2 navigation methodsand 2 sound designs

7
Lessons and Insights

• Perceptual ability
• People were able to recognize patterns of
  5-category data on a US state map, but it was
  hard.
• Designs
• Data representation form needs to fit tasks
• Interaction is important
• Navigating irregularly shaped regions is a
  challenge
• Training
• Integrated training is needed
• Observations and comments from two congenitally
  blind users show consistency with the experiment
  findings

8
New Prototype

• 3 x 3 numeric key pad to explore in 3 x 3
  sub-regions, recursively
• Absolute pointing touch pad to explore
  continuously

9
Thesis Research

• Part1 A taxonomy for interactive sonification of
  abstract data
• Interaction components
• Auditory Information Seeking Actions (AISA)
• Part2 Explore design space for
  geo-referenced abstract data
• Part3 Develop customizable user tool
  (InterSon)

10
Interaction Components
Mental representation navigation structures
Auditory interface
Auditory feedback (encoding)
AISA an interaction loop
Input device
Interaction command
Kinesthetic feedback
11
Interactive Sonification Taxonomy

• Auditory Information
• Seeking Actions Interaction
  Components

Gist Navigate, situate, search Filter by
query Select Details-on-demand Linked brushing
Abstract Object Navigation structure Input device
interaction command Auditory feedback

12
Future Work

• Refine prototype / Test
• Permit use of different maps
• Training materials
• Accessible control panels
• Dissemination
• In parallel, with support from other NSF grant,
  develop Spatial sound version with customized
  HRTF and head trackingas a longer term
  exploration of new sonification research.

13
2) Viewing Search Results with Stable Meaningful
Hierarchies

• The challenge Helping users understand search
  results

• Navigational ebay
• Known item nowell visualizing search results
• Informational visualizing search results
• Exploratory breast cancer

14
Exploratory Work Tasks

• Exploring a topic
• Writing a paper
• Developing a lesson plan
• What information is available about?
• A journalist gathering background material to
  write a series of stories on obesity in the
  United States

15
Information Seeking Model

• Understanding search results
• Gaining overviews
• Identifying unusual documents
• Meaningful context

16
SWISH Dyna-Cat
17
Honeycomb
18
Principles of search result visualization

• Provide overviewdetail display
• 100-1000 items
• Organize by meaningful, stable classifications
• Provide example documents for each category
• Use a stable visual substrate
• Arrange important text (Title, Line-in-context)
  for fast scanning/skimming
• Visually encode quantitative attributes
• Support multiple, user-controlled classifications
  and visual displays

19
Exploratory Study

• Domain Government web pages
• Two-level department/agency hierarchy
• Dept. of the Interior / National Park Service
• Legislative Branch / House of Representatives
• Motivating scenario
• Pre-computed results
• Urban Sprawl
• Breast cancer
• Alternative Energy

20
(No Transcript)
21
(No Transcript)
22
(No Transcript)
23
Exploratory Study

• Exploration sub-tasks
• What agencies provide most results
• Identify facets of topic
• Find unusual results

• Interface treatments
• 2 overview
• 1 control
• 18 subjects
• Think-aloud protocol

24
Confirmed Benefits of Overviews

• Improved accuracy
• Easier to use
• More helpful
• Users were more confident of their results
• Gaining overviews
• Finding different perspectives
• (All differences were significant)

25
Noticing Missing Results

• Users noticed areas not covered by the search
  results
• Overview conditions 9 out of 12
• Control condition 1 out of 6
• What I found informative was what didnt show
  up, which I wouldnt know if the hierarchy wasnt
  there.

26
Other Findings

• Importance of text
• Users still scanned substantial amounts of text
• Category information alone is not sufficient to
  help users gain an overview
• Expandable outliner vs. treemap
• No significant measured differences
• More preferred expandable outliner

27
Limitations of Study

• Government domain
• Narrow tasks
• Small hierarchy
• Small sample size

Recommendations

• Develop clearer categories
• Begin to integrate metatags
• Implement category browsers in search results

28
3) Narrated Demos(Recorded Animations/Videos)

• Developed series of demonstrations
• Developed set of Guidelines
• Compared across Demos

29
Our impressions

• Strong reduction in how do I questions
• Positive feedback on the demos
• Quantitative evaluation remains a challenge
• Refining guidelines
• Keeping examples of bad versions
• Will continue to create examples for new
  interfaces

30
Script Guidelines

• Base the script on a live demonstration (never on
  a written description)
• Focus on tasks(not tours of widgets or
  conceptual overviews)
• Act out the interaction (with minimum
  description) then describe results in context of
  task
• Start with a tour of main screen components
  (orient and introduce vocabulary) 5-10 sec. max
• Plan a linear sequences made of very short
  autonomous chunks (15-60 sec.)
• Map the chunks to existing online documentation
• Show text title at beginning of each chunk
• Carefully synchronize voice and visual (hard when
  alone)
• Provide duration and file size for individual
  chunk

31
Technical Guidelines

• Avoid actual video recording, use on-screen
  recording of demonstration (generates much
  smaller files)
• Make demonstration look as similar as possible to
  real interface (e.g. full screen size is better,
  crop only for readability, provide sound or
  visual effects for interaction events such as
  mouse clicks)
• Provide navigation controls if chunk is longer
  than 10sec. (Stop, Play, gtgt, ltlt, Progress
  indicator)
• Use voice of person who will be around for a
  while (i.e. plan for revisions)
• Choose the minimum connection speed to design
  for, which sets max demo duration and file size
• Provide sound transcripts and keep text
  descriptions (for deaf and blind users.)

32
Features Analysis of Demos

• Reviewed Government and other narrated demos to
  compare length, style, quality

Size Buffering Length Resembles Interface Pause Seek Time Display Voice Text Boxes Click Cues Highlighting
US Court Western District of Virginia 8100KB No 920 Yes Yes No No Yes No No No
MedlinePlus 4200KB Yes 940 Yes Yes Yes No Yes Yes No Interface
Judicial Council of California 572KB Yes 445 Yes Yes Yes No No Yes No No
LLIS 13145KB Yes 155 Yes Yes Yes Yes Yes No Visual No
Idaho Gvt software 5710KB No 131 Yes Yes Yes Yes Yes No ? Yes
EMRS software 3950KB No 1039 Yes Yes Yes Yes Yes No ? Yes
33
Features Analysis of Demos

• Reviewed Government and other narrated demos to
  compare length, style, quality
• Results coming soon ---

34
Visualization for re-identifiability

• Exposing identities from micro data
• Find vulnerable targets in micro data
• Find possible suspects in public data sources
• Link and match to re-identify
• IDFinder visualization system
• Hyunmo Kang at Census Bureau