2003-08-29

Links analysis for collaborative knowledge building
http://www.ht03.org/papers/pdfs/35.pdf


Abstract from http://www.ht03.org/papers/
"We present a research project utilizing navigation and hyperlink data to aid
collaborative knowledge building. We allow collaborators to personally organize
documents and other research resources and make references to them. We combine
their personal organizations and references to develop a unified, hierarchical
categorization of these resources. We analyze collaborators' navigations to
identify prominent research activities as well as the key documents related to
these activities. We examine prominence over time to identify research trends."


this is about an actual system used by a research community.

old system was lotus notes. purpose:
* to know what's happening in the community
* personal organisation of documents to help knowledge acquisition
* consensus organisation for collaborative knowledge building


for the new website, there's a categorisation system/association with
hyperlinks (add an item to a personal category)

and collect navigation details (from, to, when, etc)

then use link analysis for consensus knowledge building:

- Hierarchical Cluster Analysis [must look into this]

Agglomerative Clustering Dendogram:
group distance, eg centroid-based to organise groups into clusters [ -- this
looks really good, would like to do something like this. cluster Blogdex links
eg?]

- Principle Cluster Analysis (Wu et al 2002) [also must look into this. If it's
PCA there might be code I can reuse]

PCA is a form of Truncated Singular Value Decomposition (SVD) [that's another
keyword to start looking at]


[so this is a form of social software. getting common benefit for no cost, as a
side-effect of doing something you wanted to do anyway.]


in questions, the speaker says this scheme would scale up to many people, eg
350,000