generating image descriptions for smartarts - · pdf fileicchp 2016 slide 9 example: short...
TRANSCRIPT
Linz, 15.07.2016
Generating Image Descriptions for SmartArts
Presenter: Jens Bornschein
Authors: Jens Voegler, Julia Krause & Gerhard Weber
ICCHP 2016 Slide 2
Outline
• Analysis of existing materials
• What are SmartArt-Graphics?
• Markdown and SmartArt converter
• Example of a SmartArt
• Short and long description
• Concept of SmartArt converter
• Evaluation
• Conclusion & Outlook
15.07.2016
ICCHP 2016 Slide 3
Process of Lecture-Material Transcription
PowerPoint
• Used by lecturer
• Offers templates for slides
Markdown
• Add/correct content
• Easy to edit by non-HTML-experts
HTML
• Highly accessible for visual impaired readers
• Readable by AT
15.07.2016
ICCHP 2016 Slide 4
Analysis of Existing Materials (WT 2015 TUD)
• Analysis was semiautomatic
• metadata of the PDF files by a script
• PowerPoint files were counted manually if no PDF files available
15.07.2016
155166108 (1%)111 (1%)117 (1%)
648 (7%)1381 (15%)1443 (15%)
2145 (23%)3371 (35%)
0 1000 2000 3000 4000
OmniGraffle 5.4.4
Writer (LibreOffice)
Impress (LibreOffice)
Indesign (Adobe)
TextEdit (OS X)
Word (Microsoft)
PScript.dll
Keynote (OS X)
LaTeX
Other
PowerPoint (Microsoft)
pages (n= 9456)
ICCHP 2016 Slide 5
What are SmartArt-Graphics?
• Feature of Microsoft since PowerPoint 2007
• Help to illustrate processes, hierarchies, workflows, lists, …
• Easy use via templates and GUI for creation
15.07.2016
ICCHP 2016 Slide 6
Markdown and SmartArt converter
15.06.2016
Authoring Tool PowerPoint
• Headings
• Paragraphs
• Lists
• Tables
• Grouped elements
• Images
• SmartArt-graphics
Markdown converter
+
SmartArt converter
Output files in
• Markdown
• HTML
• Generated screenshot/ export
• SVG
ICCHP 2016 Slide 7
Example of a SmartArt
15.07.2016
ER-ModellRelation-Scheme
Normalized Relation-Scheme
ICCHP 2016 Slide 8
Short and Long Description
15.07.201614.07.2016
SmartArt-Graphics
Short description
Long description
Short description• Category & name of SmartArt• Number of main points• Short sentences• list of content
Long descriptionContent of short description + more detailed information
ICCHP 2016 Slide 9
Example: Short Description
Text with Placeholder:
category: #CategoryName#
subcategory: #SubcategoryName#
This graphic has #CountMainPoint# main points
#with/withoutCP# child elements.
Text with replaced Placeholders:
category: process
subcategory: running block process
This graphic has 3 main points without child elements.
The listing describes a horizontal process.
1. ER-Modell
2. Relation-Scheme
3. Normalized Relation-Scheme
15.07.2016
ICCHP 2016 Slide 10
Example: Long Description
Text with Placeholder:
category: #CategoryName#
subcategory: #SubcategoryName#
This graphic has #CountMainPoint# main points
#with/withoutCP# child elements.
<!– long description of Author –->
The #numberOfMP# main point is a #color# colored rectangle with rounded edges. The text content is #ContentOfMP#.
15.07.2016
ICCHP 2016 Slide 11
Example: Long Description
Text with replaced Placeholders:
category: process
subcategory: running block process
This graphic has 3 main points
without child elements.
The listing describes a horizontal process.
The 1st main point is a bluish colored rectangle with rounded edges. The text content is ER-Model.
The 2nd main point is a greenish colored rectangle with rounded edges. The text content is Relation-Scheme.
The 3rd main point is a orange colored rectangle with rounded edges. The text content is Relation-Scheme.
15.07.2016
ICCHP 2016 Slide 12
Concept of SmartArt converter
15.07.2016
1. Validation (semiautomatic)
a. Are all required descriptions available?
b. User Feedback – enter missing descriptions
2. Creation of descriptions (automatic)
a. Filtering of all SmartArt-graphics
b. Replacement of placeholder
c. Export SmartArt-graphics as pictures
3. Export of descriptions
ICCHP 2016 Slide 13
Evaluation
• 4 blind subjects
• 3 Phase Evaluation
1. Questionnaire about current situation
2. Task: rebuild 5 SmartArt-graphics
1. Automatic description
2. Rebuild
3. Manual Description
4. Refine
3. Feedback about the quality of descriptions
15.07.2016
ICCHP 2016 Slide 14
Evaluation Pre-Questionnaire Results
• 3 of 4 have no knowledge about SmartArt
• 4 of 4 do not read alternative description during a lecture
• 4 of 4 evaluate the quality of conventional descriptions from “adequate” to “good”
15.07.2016
Very
important
Less
importantneutral
rather
unimportantunimportant
Structure of the description x x x x
Textual content x x x x
Description of the relation
between the elements
x x x x
Color information x x x x
Description of the geometric
shapes
x x x x
Description of the foreground x x x x
Description of the background x x x x
ICCHP 2016 Slide 15
Evaluation Example 1
15.07.2016
Abhängige Variable:
Ängstlichkeit
Stör‐Variablen (Geschlecht, Alter,
Untersuchungslärm…)
Unabhängige Variable: Intelligenz
After automatic desc. After refinement (man. desc.)
ICCHP 2016 Slide 16
Evaluation Example 2
15.07.2016
alle Relationen
1. NF
2. NF
3. NF
BCNF
4. NF
5. NF
After refinement (man. desc.)After automatic desc.
ICCHP 2016 Slide 17
Evaluation Example 3
15.07.2016
ER-Modell Relationen-SchemaNormalisiertes
Relationenschema
No refinement needed
ICCHP 2016 Slide 18
Evaluation Post-Questionnaire Results
• 3 of 4 - short descriptions would be read during lecture
• 2 of 4 - structure is helpful
• 2 of 4 - prefer the manual description
15.07.2016
ICCHP 2016 Slide 19
Advantages
• Changeable descriptive texts
• Unification
• “Pattern” recognition (mapping the title to the layout)
• Rapid creation
• Faster access for students
• Less effort (compared to manual desc.)
15.07.2016
ICCHP 2016 Slide 20
Conclusion & Outlook
• Improve the use of articles, plural and singular of words
• Iterative process of producing description templates in cooperation with blind students
• Development of a concept of sharing created SmartArt descriptions
15.07.2016
ICCHP 2016 Slide 21
Contact the Author
Dipl.-Medieninf. Jens Voegler
Technical University of DresdenFaculty of Computer ScienceInstitute of Applied Computer ScienceChair Human-Computer Interaction01062 DresdenGermany
phone: +49 (351) 463-42024fax: +49 (351) 463-38491E-Mail: [email protected]
15.07.2016
ICCHP 2016 Slide 22
Thanks for your attention
15.07.2016