Text Scanner

1
Text Scanner

• Alex Krstic
• Kelly Van Busum
• Suzanne Vogel

2
Outline

• Problem Specification
• General Approaches
• Examples
• Our Approach
• Semester Goals
• Current Status

3
Problem Specification

• Observations
• Hearing text read aloud is very SLOW
• Hearing books on tape
• Sometimes you dont need to hear every word
• Hearing headlines on CNN
• Some experts think hearing every word
  (subvocalization) decreases comprehension.
• What is a BufferedReader?

4
How Can We Help?

• By understanding how we process what we hear
• For example
• Can we leave some sounds (words) out?
• Jeetyet?
• How much can we hear at once?
• Speed reading reading in chunks, horizontally
  and vertically

5
And How Do We Do That?

• Learning Disabilities
• Why do some people process information better
  than others?
• Brain function
• Visual vs. Audio
• Speed Reading
• How do we speed up visual processing?

Trying to figure out how this program should
work!
6
General approach

• Segmentation (no loss of info)
• - splitting document
• Filtering (possible loss of info)
• - levels of detail/hierarchy

7
General approach (contd)

• Augmentation
• - spatial sound
• cocktail party effect
• - earcons
• Compound
• - audio icons

8
Navigation

• Scanning (retains content)
• - removing redundant info
• - ex. spaces, word overlaps
• Skimming (lose content)
• - removing (non-)redundant info
• - ex. entire segments, comb sampling

9
LOD Example SpeechSkimmer
(time compression)
10
Why use time compression?

• Speed limit 2x normal
• - normal speed 130 200 wpm
• Time compression better for gt2x normal

11
Navigation Example ASTER

• Based on the predefined format,
• e.g. math, markup languages
• - reading formatted documents in hierarchical
  order
• - hyperlinks and bookmarks

12
Our Approach

• Goal Balance comprehension and speed
• Remove common words
• Remove redundant phonemes
• Hierarchy for navigation
• Spatial sound

13
Current Status

• Read about theory and prior work
• Began implementation (Python)
• GUI (to be improved)
• Tokenizing text files
• Experimentation with text (quick-n-dirty)
• Meet 2x/week

14
Acknowledgement

• Pete Parente
• Pointed us to programming resources (BATS
  wxPython, Python Numeric 22.0, Win32 libraries)
• Gave us Python sample code for speech synthesis
  and spatial sound
• Thank you! What would we do otherwise?