a document skimmer overcoming the soda-straw effect alex krstic kelly van busum suzanne vogel
Post on 21-Dec-2015
221 views
TRANSCRIPT
A Document SkimmerA Document SkimmerOvercoming the soda-straw effectOvercoming the soda-straw effect
Alex KrsticAlex Krstic
Kelly Van BusumKelly Van Busum
Suzanne VogelSuzanne Vogel
OutlineOutline
Problem Overview Problem Overview Prior Work (briefly)Prior Work (briefly) Our WorkOur Work DemoDemo StudyStudy Follow upFollow up
Overview: ProblemOverview: Problem
Listening is slower than reading, but Listening is slower than reading, but speeding up decreases speeding up decreases comprehensioncomprehension Speed up only by increasing reading Speed up only by increasing reading
rate, with NO scanning or skimmingrate, with NO scanning or skimming Skip ahead only by one line or one pageSkip ahead only by one line or one page
Overview: GoalOverview: Goal
Identify features to increase speedIdentify features to increase speed Enable the user to adjust these Enable the user to adjust these
featuresfeatures Trade off speed and comprehensionTrade off speed and comprehension
Prior Work: FeaturesPrior Work: Features
Scan at levels of detail (LODs)Scan at levels of detail (LODs) Speech SkimmerSpeech Skimmer [1] & [1] & AsterAster [2] [2]
Skip 1 segment within a levelSkip 1 segment within a level Speech SkimmerSpeech Skimmer [1] [1]
RefsRefs1.1. Speech SkimmerSpeech Skimmer (Arons, 1993) (Arons, 1993)2.2. AsterAster (Raman, 1994) (Raman, 1994)
Prior Work: ImplementationPrior Work: Implementation
Segment document, semanticallySegment document, semantically Speech divisions: Long pauses [1]Speech divisions: Long pauses [1] Text divisions: Structure boundaries [2]Text divisions: Structure boundaries [2]
Filter out words or sounds within Filter out words or sounds within segmentssegments Spaces [1]Spaces [1] Latter Latter P P number of words or seconds [1] number of words or seconds [1] Detailed (lower-level) info [2]Detailed (lower-level) info [2]
Our Work: FeaturesOur Work: Features
HierarchyHierarchy
Dropping Words/PhonemesDropping Words/Phonemes
Spatial SoundSpatial Sound
Our Work: LOD HierarchyOur Work: LOD Hierarchy
Our Work: Dropping Our Work: Dropping Words/SoundsWords/Sounds
Dropping common wordsDropping common words
Change text to phonemesChange text to phonemes toz, suhntoz, suhn
Remove phonemes without lexical stressRemove phonemes without lexical stress computing computing mpyootng mpyootng
Blending phonemes (Drop spaces)Blending phonemes (Drop spaces) what up what up whuhtuhp whuhtuhp
Our Work: Spatial SoundOur Work: Spatial Sound
Hearing more than one sound source Hearing more than one sound source at the same timeat the same time
2, 3 or 42, 3 or 4
Each source plays different segments Each source plays different segments of the fileof the file
Some sources dominant over the Some sources dominant over the othersothers
Spatial orientationSpatial orientation
Our Work: ScreenshotOur Work: Screenshot
Copyright 2003, ASK (Alex, Suzanne, Kelly)
User EvaluationsUser Evaluations
3 informal, 4 systematic3 informal, 4 systematic
Asked questions, navigate to answerAsked questions, navigate to answer
Hear text in various forms, then Hear text in various forms, then asked questionsasked questions
User Evaluations, 2User Evaluations, 2
HierarchyHierarchy Difficult to explain “hierarchy concept”, underusedDifficult to explain “hierarchy concept”, underused
Sound (Word) RemovalSound (Word) Removal Removing common words was liked (29% of words)Removing common words was liked (29% of words) Either really liked or hated phonemes (29%, 10%) Either really liked or hated phonemes (29%, 10%)
Spatial SoundSpatial Sound 2 sounds worked ok, 3 or more didn’t2 sounds worked ok, 3 or more didn’t
*Lots of different perspectives!*Lots of different perspectives!
New Questions…New Questions…
How much does voice selection matter?How much does voice selection matter?
How much would training help?How much would training help?
What is the relationship between phonemes and What is the relationship between phonemes and speed?speed?
What is the role of prior knowledge? What is the role of prior knowledge?
How does this relate to Ctrl-F?How does this relate to Ctrl-F?
AcknowledgementsAcknowledgements
Peter ParentePeter Parente Pointed us to programming resources Pointed us to programming resources
(BATS; wxPython, Python Numeric 22.0, (BATS; wxPython, Python Numeric 22.0, Win32 libraries)Win32 libraries)
Gave us Python sample code for speech Gave us Python sample code for speech synthesis and spatial soundsynthesis and spatial sound
Experiment participantsExperiment participants (Informed consent requires (Informed consent requires
confidentiality)confidentiality)
Programming ResourcesProgramming Resources
BATS NCDemo – BATS NCDemo – http://www.sourceforge.nethttp://www.sourceforge.net OpenAL.dll, MSVRTD.dll, pyTTS.py, pyOpenAL.py OpenAL.dll, MSVRTD.dll, pyTTS.py, pyOpenAL.py
(I think)(I think) Python – Python – http://http://www.python.org/www.python.org/ Win32 library for Python – Win32 library for Python –
http://starship.python.net/crew/mhammond/http://starship.python.net/crew/mhammond/ Python Numeric 22.0 library – Python Numeric 22.0 library –
http://www.pfdubois.com/numpy/http://www.pfdubois.com/numpy/ wxPython GUI library – wxPython GUI library –
http://www.wxpython.org/http://www.wxpython.org/