Communications of the ACM
Time Travel Is Now Possible
The Computer-Supported Collaborative Work (CSCW) conference kicked off this morning and as one of three CACM bloggers at the conference I have the privilege of reporting about some of the fantastic research being presented.
The keynote is being covered by my colleague, Michael Bernstein, but I will be reporting on two of the sessions I attended this afternoon. One session was on analyzing interactions, which dealt primarily with methods for measuring interactions, and the second was entitled meeting in the middle, which dealt primarily with ways of improving group dynamics and idea generation. While all of the presentations in both sessions were interesting in their own right, I have selected one presentation from each to provide a taste of the innovative research being presented at CSCW.
From speaking to attendees who were at other sessions, I know there seemed to be particular interest in the first of the papers presented in analyzing interaction, Psycholinguistic Network Analysis (the title given in the talk, however, was Social Language Network Analysis). It should come as no surprise that this work is not yet fully mature; however, the use of linguistic analysis in order to aid social networks has a lot of potential. The general gist of this research: Andrew Scholand (Sandia National Laboratories), Yla Tausczik (University of Texas at Austin) , and James Pennebaker (University of Texas at Austin) first used linguistic analysis to determine the general strength of relationships and then combined that with social network analysis to determine which individuals are friends. The talk generated good discussion during the question and answer session for ways to expand the work in the future, and I hope we’ll see more from the researchers on this topic.
Listed as among the best of CSCW, CatchUp announced to the audience that "time travel is possible (but it’s currently limited to catching up on meetings)." This project aims to solve the problem of every academic—catching up on a meeting, after having arrived late, without having to derail the meeting through a recap that everybody must endure. The solution offered by University of Sheffield's Simon Tucker, Ofer Bergman, Anand Ramamoorthy, and Steve Whittaker comes from a summarization process they refer to as "gisting."
The portion of a meeting an individual has missed is recorded and run through an automatic recognition system (AMIDA), which has a 35% error rate for conversational voice recognition, and then identifies the importance of information using the TFxIDF process, which looks at frequency of information in the current transcript and in previous, similar transcripts. It then aurally replays these key segments to the user at a defined rate of how much to include/exclude.
In a multiple-choice test of information presented at the meeting, the experimental group was found to perform significantly better than the group that missed the meeting without being able to travel through time. However, the catch-up process does not come free; in the eight-minute meeting used for the experiment where half of the meeting was missed by study participants, it took two minutes and forty seconds to recap the meeting (experimental participants were able to only attend to the CatchUp auditory stream and the final minute and two seconds of the meeting). Audience members suggested several possible solutions to reduce this high cost to catching up on the content.
That’s it for day one from me. I am told that one of my co-bloggers will write about the demo and poster sessions—some really great projects showcased there as well. I’m looking forward to filling everybody in on more papers and notes tomorrow when I plan to attend CSCW Horizons and Groupware Technologies. Then I am student volunteer for an evening session so I will take notes (if the opportunity arises). If there are any papers from those sessions that anybody wants me to give a quick recap, please leave a note in the comments, and I will try to cover it.
---
About the Author
Michael A. Oren is a Ph.D. student at Iowa State University co-majoring in Human-Computer Interaction and Sociology (where his emphasis is on sociological theory). He has a B.A. in computer science and English (creative) writing from DePauw University and a M.S. in Human-Computer Interaction from Iowa State University. This is his first time attending CSCW, but he has previously attended other ACM conferences including SIGCSE, CHI, and ASSETS. He is currently reading Sociologial Insight by Randall Collins. Although it is a fairly introductory text from a sociological theory point of view, it does have an interesting (if dated, with the last edition released in 1992) chapter on how sociological theory can aid artificial intelligence researchers.
No entries found