Information about http://ils.unc.edu/~march/journalism_april08.pdf
…
Tags: ch 4, colloquium series, conversation school, digital libraries, dv projects, election case, library science, mary junck, north carolina at chapel, north carolina at chapel hill, public expression, research colloquium, school of information, school of journalism, science university, social interaction, theoretical view, university of north carolina, university of north carolina at chapel hill, youtube,Pages: 6
Language: english
Created: Thu Apr 17 12:03:52 2008
Display cached document
Page 1
Page 2
Page 3
Page 4
Page 5
Page 6
Outline
Digital Video: From Thesis: Digital Video blurs traditional information and
communication boundaries
Digital Libraries to Three ongoing DV projects
Social Interaction Open Video DL
· Accessible, reusable files
· Surrogation as key R&D challenge
The Mary Junck Research Colloquium Series · User studies on surrogate effects
School of Journalism and Mass Communication · The question of channel synchronicity
University of North Carolina at Chapel Hill VidArch and preserving context
· YouTube and 2008 Election case
April 17, 2008
UNC YouTube Channel: issues and policies
Gary Marchionini From retrieval to (public) expression and conversation
School of Information and Library Science
University of North Carolina at Chapel Hill
march@ils.unc.edu 2 Gary Marchionini, UNC-CH 4/17/2008
www.ils.unc.edu/~march
Digital Video Status Theoretical View
Digital video a burgeoning DL challenge: YouTube phenomenon Digital video is crafted expression
(fall 2007: 65K new videos/day; 20TB/mo; 100M views/day) Multiple channels (analog and digital)
Substantial research activity on storage, retrieval from engineering Visceral as well as intellectual effects (analog and digital)
perspective (see IEEE, ACM MM) A descendant of film but with potential dynamics/behavior (digital)--changes over time, every
time
Many large-scale DLs and services
InforMedia, Fischlar, ECHO, Internet Archive , Open Video, Digital Libraries are journeys (learning
public.tv, researchchannel environments) rather than destinations for patrons
Most attention on system/collection building rather than services and librarians
Commercial attention on system and management Beyond libraries as repositories to sharium
IBM, MERL, Microsoft, Artesia, Virage, WFMY locally
NIST TREC Video Track for retrieval evaluation
Open Video deals with reusable (open) video
Advances on capture, critical need for reuse tools
objects
A journey toward new forms of expression and reflections on history
Portability advances (e.g., wifi, iPhone) What do you do with 24/7 feeds of video from every street corner in Manhattan?
On the cusp of `natural'
3 Gary Marchionini, UNC-CH 4/17/2008 4 Gary Marchionini, UNC-CH 4/17/2008
Open Video Vision/Contributions
http://open-video.org
Background & Status
An open repository of video files that can be re-used in a variety Begun 1995 with colleagues at UMD & BCPS; current instance at
of ways by the education and research communities UNC initiated in 1999
Encourages contributions Funding: NSF# IIS-0099538 1999-2004; NSF IIS 0455970 2006-
07; Library of Congress NDIIPP (2007-08); IBM (2007); Google
A testbed for interactive interfaces (2007)
An easy to use DL based upon the agile views interface design Collaborators/Contributors: I2-DSI, ibiblio, CMU, UMD, Prelinger
framework Archive, Internet Archive, NASA, ACM
Multiple, cascading, easy to control views (pre, over, re, ~4000+ video segments
shared, peripheral) ~40000 unique visitors per month
Views based upon empirically validated surrogates ~1.8M hits/month
An environment for building theory of human information MPEG-1, MPEG-2, MPEG-4, QT
interaction OAI provider
A set of methods and metrics that reveal how people understand Ongoing user studies of surrogation
digital video through surrogates
5 Gary Marchionini, UNC-CH 4/17/2008 6 Gary Marchionini, UNC-CH 4/17/2008
1
What is (was) a Surrogate? The Blur
The (relatively) Neat Past and the Very Scruffy
Condensed representation for human consumption Present
constructed to stand for an information object Blurring the `levels of representation' model of information
(primary-secondary-tertiary-n-ary)
Information compressions The metadata--surrogate continuum within the levels
Surrogates of representation continua
Enable decision-making by presenting search results in a uniform way Metadata region mainly for retrieval
Support sense making and incidental learning
Save human time (compaction)
Metadata region mainly for and by machines (semantic
web)
Save network capacity and system resources
· Automatic metadata generation advances
Examples · Implicit links and mining of interactions as metadata
Abstract, gloss, summary Surrogate region mainly for sense making
Title, bibliographic record
Preview, snippet
Surrogate region mainly for and by people
Profile · Professional abstracting
Logo · Social tagging and annotations/links as surrogates
Your avatar
7 Gary Marchionini, UNC-CH 4/17/2008 8 Gary Marchionini, UNC-CH 4/17/2008
Representation and the Digital Blur
Physical
World
Mental
Digital Video Surrogates
Information Classes
Representation
Object
Level 1 Textual
Recorded Ephemeral
Visual
Representation Audio
Level 2
metadata surrogates Cost benefit analysis: maximize `meaning'
per unit time
World
Physical Mental Transmission time
Compaction rate
Digital Information
Representation Object Cognitive processing time
metadata annotations
Performance vs. Preference
surrogates
9 Gary Marchionini, UNC-CH 4/17/2008 10 Gary Marchionini, UNC-CH 4/17/2008
Research Framework Surrogates Examined
Storyboard with text keywords (20-36 per board@ 500 ms)
Storyboard with audio keywords
GOALS
learning, work, entertainment Slide show with text keywords (250ms repeated once)
Slide show with audio keywords
EFFORT TASKS OUTCOMES
Fast forwards 32X, 64X, 128X, 256X
TIME select video for viewing PERFORMANCE
time spent searching select scene for viewing retrieval (precision, recall) Poster frames (1-3)
and viewing results copy and use scenes recognition (objects, action)
VIDEO copy and use frames INDIVIDUAL gist comprehension Real time clips/excerpts (7 sec)
MENTAL LOAD CHARACTERISTICS other tasks? CHARACTERISTICS (linguistic, visual)
perceptual load Text (title, keywords, etc.)
cognitive load genre: documentary, domain experience
SATISFACTION
narrative
topic: literal, figurative
video experience
cultural experience
perceived usefulness Visual features (e.g., in/out, people, etc.)
PHYSICAL LOAD perceived ease of use
amount of muscle
style: visual, audio,
textual, place
SURROGATES,
AGILE VIEWS
computer experience
info seeking experience
flow Spoken descriptions
movement user satisfaction
display controls
metacognitive abilities
demographics
Spoken keywords
keywords
storyboard w/ text, audio
Combined visual (storyboard, fast forward) and spoken
slide show w/ text, audio
fast forward w/ audio
(descriptions, keywords)
poster frames
11 Gary Marchionini, UNC-CH 4/17/2008 12 Gary Marchionini, UNC-CH 4/17/2008
2
Tasks User Studies
Text Still Image Moving Image Audio Qualitative Comparison of Surrogates (Spring 02, ECDL 02)
Fast Forwards (Fall 02, JCDL 03)
Recognition/ Object selection (text) Object selection Excerpt selection
Select Spoken Text or Pictures (Spring 03, CIVR 03)
Selection Keyword selection (graphical)
Description selection Keyframe selection
Description Narrativity (CHI 02, ASIST 03)
Select Spoken
Title selection
Keyword Shared views and History Views (Geisler dissertation)
TREC evaluation (Spring/summer 03; 05)
Generative Gist writing Visual gist
Inference (free text) determination ViSOR (Gruss Master's paper)
Look vs Read (Hughes Master's paper)
Video relevance (CHI 05; ASIST04; Yang dissertation)
Metrics
Cognitive load (Mu dissertation)
Accuracy Teachers using video (Brown dissertation)
Confidence Spoken Audio and Storyboards (CHI 07)
Time to complete Spoken Audio and Fast Forwards (current)
Usefulness, usability, engagement, enjoyment, preferences
13 Gary Marchionini, UNC-CH 4/17/2008 14 Gary Marchionini, UNC-CH 4/17/2008
Audio Surrogates
Spoken descriptions, summaries,
keywords
Visual displays of audio signals
Audio skims (excerpts)
Compressed speech
Parallel streams (cocktail party effect)
15 Gary Marchionini, UNC-CH 4/17/2008 16 Gary Marchionini, UNC-CH 4/17/2008
Recent Study (CHI 07) Synchronicity
36 participants, within subjects design used audio-only Coordinated media channels lead to
(spoken descriptions), visual-only (storyboards), and
combined surrogates to do 5 kinds of recognition and gist better understanding, retention, and
tasks. satisfaction
Accuracy, time to view, time to complete task, suite of
affective measures What about multi-channel surrogates?
Statistically reliable differences on 3 of 5 accuracy tasks,
time to view, and most affective measures. Combined Assume surrogate channels should also be
generally better and preferred, audio almost as good as coordinated?
combined, visual alone faster to consume but no time Perhaps more sense making possible if sampling
penalties for audio and combined on task completion. across different channels and integrating in the head
Implications at consumption time rather than pre-coordination at
Add audio surrogates indexing time?
Use audio in small form-factor devices
Audio and visual quality important We have initiated a series of studies
Synchronizing different channels in surrogates may not be
necessary
User controlled tradeoffs: time, satisfaction, performance
17 Gary Marchionini, UNC-CH 4/17/2008 18 Gary Marchionini, UNC-CH 4/17/2008
3
Tradeoffs Video Preservation
(VidArch) Project
http://ils.unc.edu/vidarch
Visual
channels
What kind and how much context to preserve?
National Digital Information Infrastructure Preservation Program
OR? (NDIIPP) funding via NSF and LoC.
Audio
Focus on specific topics
channels 2008 Presidential campaign (15K May 07-present)
Energy, truth commissions, health, pandemics
= most salient samples Harvest video, metadata, and activity from YouTube; use API to
User-centered integration query rather than crawl
Pre-processed integration
yields less cognitive load. (constructivist). More cognitive Create Curator's tools and services
Less sense making and load. Better sense making and Fundamental DL issue of content/metadata/context boundaries
retention? retention due to active in WWW objects
participation and ?better?
Information samples
19 Gary Marchionini, UNC-CH 4/17/2008 20 Gary Marchionini, UNC-CH 4/17/2008
Challenges of Extending
Extend Documentation Documentation Strategies
Strategies to Web to Web Videos
Power of the masses to produce
Potentially too many for hand-crafted,
documentation of a subject or issue
artisan approach.
(folksonomy).
Web-based materials can have a strong Ephemerality here today; gone
impact on society, including phenomena tomorrow.
such as voting behavior. Variable quality and relevance.
Democratizes collecting strategies; more General lack of metadata.
than network news or campaign materials Unclear provenance and authenticity.
being collected.
Lack of contextualizing information.
Materials never before created or collected.
21 Gary Marchionini, UNC-CH 4/17/2008 22 Gary Marchionini, UNC-CH 4/17/2008
Video Harvesting from Election 2008 Collecting
YouTube Scenario
Curator of Hillary Clinton's campaign.
Direct feed of materials from Clinton's staff and
Democratic Party.
Press releases, video, interviews, Face Book, etc.
Wide variety of traditional media newspapers, TV,
radio.
Now wide variety of bottom-up materials, including You
Tube videos
"Official" CNN debate videos, reactions, etc.
23 Gary Marchionini, UNC-CH 4/17/2008 24 Gary Marchionini, UNC-CH 4/17/2008
4
Video Harvesting from Overview of the collection
YouTube for Election 2008 (as of 04/17/2008)
56 queries (6 general and 50 names) Crawls = 273
Use You Tube APIs, screen scraping and other tools to collect Unique videos = 17,862
videos and context
Total videos = 19,570
Crawl everyday (almost) since May 07
Video files = 181 GB
Get top 100 results for each query
Total views = 496,581,313
Collect more than 20 attributes (including all the comments)
Total comments = 3,017,625
Download flash videos
Total ratings = 2,847,427
Compare to blog postings Total honors = 547
Capra, R., et al., (in press). Selection and Context Scoping for Digital Video Collections: An Investigation of YouTube and
Blogs. ACM/IEEE Joint Conference on Digital Libraries (June 2008).
Also see http://ils.unc.edu/vidarch
25 Gary Marchionini, UNC-CH 4/17/2008 26 Gary Marchionini, UNC-CH 4/17/2008
Implications to Date UNC YouTube Channel
YouTube is as much a conversation as an http://youtube.com/uncchappelhill
information source
Comments and responses Content
· Textual
Information in Life Series
· Video
Layers of video representation that are · Lectures (SILS, ibiblio, Public Health)
strongly culture dependent · Interviews with UNC Faculty/staff
· Video allusions (e.g., McCain's Obama girls,
mama, etc.; Vote different) Carolina Week playlist
The Internet is not quite mainstream Global Health playlist
· Ron Paul in blogosphere and YouTube
News and Publicity
27 Gary Marchionini, UNC-CH 4/17/2008 28 Gary Marchionini, UNC-CH 4/17/2008
Policy Issues Implications
Collection Development
Campus gatekeeping The Medium get attention, attention
Instructor ownership
brings new requirements for response
Intellectual Property and Reuse
Sturm's viral lecture on storytelling
CC
Incidentals Carson memorial comments
Management New opportunities and challenges for
Content development teaching and learning
Channel freshness
Blowback
Marchionini, G. (in press). Digital Video Policy and Practice in Higher Education: From Gatekeeping to Viral Lectures.
Educational Technology
http://youtube.com/uncchappelhill
29 Gary Marchionini, UNC-CH 4/17/2008 30 Gary Marchionini, UNC-CH 4/17/2008
5
Summary Implications: Q&A
The beginning of an information and
communication paradigm shift Thanks for your attention
Lots of content to reuse
· Professional to web/security cam
Capture everything mentality
· Public/private blur
Non-textual tags and annotations Thanks also to NSF, Library of Congress, NASA,
· Primary/n-ary blur
Microsoft, IBM, and Google for partial support of this
Conversation interjections work
Create/find/share blur
31 Gary Marchionini, UNC-CH 4/17/2008 32 Gary Marchionini, UNC-CH 4/17/2008
6