http://detentionblockaa32.blogspot.com/2011/04/paper-reading-23-natural-language.html
Facilitating Exploratory Search by Model-Based Navigational Cues
Wai-Tat Fu, Thomas G. Kannampallil, and Ruogu Kan
University of Illinois
Presented at IUI’10, February 7–10, 2010, Hong Kong, China
Summary
The authors of this paper wrote it and built a simulator to test the notion that unstructured social tagging may cause difficulties for searchers. The hypothesis being challenged is based on the notion that casual tagging will eventually become an incoherent mess of tags. The counter hypothesis is that the tagging isn't as random as thought, and will instead follow cohesively from whichever tags are posted earliest. That is, early tagging heavily influences the tagging of later users.
The Semantic Imitation Model was designed to simulate the actions of expert and novice users across a document space assembled for the study. The results of the simulation did seem to indicate that convergence is experienced.
from the paper
Discussion
I have to be honest, I don't think much of this paper. First, the problem statement being challenged is very weak - the suggestion that tagging will go all over the place and become useless is very counter intuitive, and it almost feels like the authors built a strawman to have something to challenge.
I really don't understand why they ran simulations rather than finding some actual users to do the study for them...if you build searcher simulators to model searched, they will be based on your understanding of how searchers - even if you don't intend them to - and will tend towards validating your understanding of the whole system.
It's possible that some of the above criticism is invalid, and I simply misunderstood. In that case, the paper's flaw is rather a failure to communicate effectively. Future work : do it again, with people this time.
In the interest of completeness : significant because due to the prevalence of searching any improved understanding thereof is quite useful.
I also read this paper and completely agree with you. Their study just seemed a little forced. Also, they should have mentioned Google's image tagging app or something that shows real world results.
ReplyDeleteIt seems like we are reading the same papers and I agree with you. This paper did not do much for me. It seems to me that we have been reading multiple papers with the same idea of navigational cues, social tags. I found this paper boring and like you mentioned, they should have used actual people in their testing.
ReplyDeleteYou said it. Tagging systems are a troller's paradise.
ReplyDelete