Collaboration and Education Group Jonathan Grudin Microsoft Research
Collaboration and Education Group Formed about 12 months ago Mission: To explore novel technologies and applications that enhance collaboration and education / training Current work focuses on streaming media Research model Evaluation: Laboratory and Field Studies Build Prototype Evaluation / Publication Refine Prototype Product Impact
Technology and Education Two broad facets: Technology for improved content deep models of subject matter and student active exploration of subject (simulations) relate to students context/environment (situated learning) MOSTLY DOMAIN DEPENDENT Technology infrastructure for: course and student management content creation delivery / distribution collaboration MOSTLY DOMAIN INDEPENDENT Both aspects are important and complementary
Technology Adoption Phases Phase-1: digital version of non-digital process Phase-2: value-added features appear in digital version Phase-3: process and technology re-design
Why Consider Multimedia? Network, processor, memory capability changing quickly Reasoning about exponential growth Simultaneous emergence of live and on-demand capability Shift in the definition of scholarship
Ongoing Projects MSTE and MURL: Online Seminars Time Compression, Skimming, Indexing, Browsing MRAS: Multimedia Annotations and Authoring Flatland: Telepresentations
MSTE Presentations Logs of ~10,000 sessions by over 2000 users Some results: On-demand audience about 40% of live audience 60% < 5 minutes Viewers jump around video Initial portions much more likely to be watched Presentations will be designed differently in future Present key messages early in talk Present key messages early in slide Use meaningful slide titles Reveal talk structure in slide titles Consider post-processing talk for on-line viewers
Analysis of Online Presentation Viewing Logs of ~10,000 sessions by over 2000 users Some results: On-demand audience about 40% of live audience 60% < 5 minutes Viewers jump around video Initial portions much more likely to be watched Presentations will be designed differently in future Present key messages early in talk Present key messages early in slide Use meaningful slide titles Reveal talk structure in slide titles Consider post-processing talk for on-line viewers
Ongoing Projects MSTE and MURL: Online Seminars Time Compression, Skimming, Indexing, Browsing MRAS: Multimedia Annotations and Authoring Flatland: Telepresentations
Time Compression, Skimming, Indexing While text documents are easy to skim, that is not true for audio-video Ability to skim can be a key advantage of web-video time-compression: up to ~2-fold; nothing thrown away skimming: > 2-fold; some content thrown away indexing: adding navigable structure Also useful in “live” broadcast scenarios e.g., late joiners can catch up to live talk
Time Compression: Synchronized Audio and Video To preserve pitch: throw away portion of each 100ms chunk, then stitch together Basic signal processing well known, but several systems issues Results of lab studies: People choose ~1.4 speed, don’t adjust much They like it “I think it will become a necessity… Once people have experienced it they will never want to go back. Makes viewing long videos much, much easier.” Comprehension may go up
Time-Compression Demo
Skimming: Compression Goes Nonlinear To beat 2x speedup, must throw away content Sources of information audio: pauses, intonation, speech-to-text and NLP video: scene changes other: slide-changes, previous viewers’ patterns Lab studies of 4x-5x speedup Viewers learn from automatic summaries Viewers like and learn more when author-edited Perception of quality increases over time Mixed-initiative summarization is promising
Indexing Vanilla video provides no structure for navigation Indexing provides navigable structure; examples: textual table of contents (slide titles) video shots / scenes speech-to-text => NLP => topic detection
Ongoing Projects MSTE and MURL: Online Seminars Time Compression, Skimming, Indexing, Browsing MRAS: Multimedia Annotations and Authoring Flatland: Telepresentations
Multimedia / Temporal Annotations Motivating scenarios: a virtual university all students are remote, asynchronously watching lecture videos a standard university making better use of in-class time Temporal annotations: annotations associated with streaming media each annotation is linked to the media time-line annotations stored separately from the media files
Ability to annotate can add significant value shared notes for asynchronous collaboration question-answers linked to a streaming-video lecture archived feedback for the instructor personal notes on audio-video found on the web personal/shared table of contents; summarizations annotations may be computer generated use speech-to-text providing search and seek ability captured strokes from electronic white-board captured questions, slide-flips, from “live” broadcast ...
Results from Preliminary User Studies Personal note-taking study (MRAS vs. Paper) similar # of notes (~1 / minute) positioning: none in paper; ~10-15s later in MRAS all subjects preferred MRAS (although more time), and thought more useful for future reference Shared notes study text preferred to audio 14/18 stated more participation than in “live” session auto-tracking particularly useful
Currrent Work MSTE class to use MRAS and recorded lectures Can we increase instructor productivity? Can we emulate live-classroom discussion / community formation in an asynchronous environment using MRAS?
Ongoing Projects MSTE and MURL: Online Seminars Time Compression, Skimming, Indexing, Browsing MRAS: Multimedia Annotations and Authoring Flatland: Telepresentations
Flatland Tele-presentation System Joint project with the Virtual Worlds Group Flexible architecture for distributed collaborative applications Target scenarios: presentations to remote audience online conferences distributed tutored-video-instruction ...
The Flatland Project
Do We Need to Sacrifice Quality? The goal is to improve it Stanford Tutored Video Instruction (TVI) Process: video tapes of un-rehearsed live lectures small group of students watch along with a para- professional tutor Results from All MSEE: 1800 students, avg. GPA 3.40 TVI-MSEE: 89 students, avg. GPA 3.62 Similar observations recently for D-TVI version
Stanford TVI Experiments: 10/73 - 3/74 remote TVI students with tutor do best it helped “at-risk” students even more Source: J.F. Gibbons, et al. Science, Vol. 195, No. 4283, 18 March 1977
Flatland Experiences Initial use in 3 multi-session MSTE classes Presentations from desktop to remote audience Students: Liked the convenience Liked ability to multitask Did not think learning suffered Instructors: Missed familiar sources of feedback Comfort level rose over time for 2 of 3 Overall: Lack of awareness of others a key problem
Issues Being Explored Creating presence and awareness representing attendees; gaze; activity level;... Providing for interactivity; protocols for online talks types of widgets; floor control; multiple back channels Complexity of interface for speaker / audience use of channels over time; different physical contexts; … Capture and replay of tele-presentations capture “all” activity; time-compression; annotations
Activity Surrounding Teaching/Learning Pre-authoring Slides, web notes, reference material, exercises, … Content delivery Synchronous delivery to local/remote audience Archived for on-demand audience and review On-demand access by students Watch content; personal notes; TOC; index; … Discussion around content Synchronous: small group; one-on-one Asynchronous Post-lecture work by instructor / tutor Answer questions; discussions; feedback & redesign; … Student evaluation …
Concluding Remarks Key drivers of change market needs technology Key new directions learner-centric asynchronous; small-group synchronous Key challenges concrete studies to indicate effectiveness technology/products taking value beyond cost business model and bootstrapping issues
For More Information:
Watching Behavior Within a Session Nth minute into the talk User count