effjay

Daniel Stephan
Computer Science Student & Software Developer
Bremen/Germany
see also http://www.danielstephan.de

Interests & links:

• IM Expert Search: Snipsnap goes peer to peer?
• http://www.darrenzieger.com/peertopeercentral/index.html : Some nice articles about what peer to peer means or should be.
• http://www.vistaportal.com/ : A company with a plan. Maybe a source of inspiration.

First thoughts:

An expert could classify his expertise

• explicitly by selecting a number of topics or
• indirectly by classifying his documents

How should the topics be modelled?

• centrally with a system-wide taxonomy
• locally with a specific taxonomy

If we do it lokally (the p2p system would promote it...) we have to find a way to integrate the different taxonomies on the different clients. Maybe it is easy, because the topics are written in the same language and thus have to be the same. And the technical terms are the same anyway. So there should not be confusion in most situations. And the other situations can be ignored :)

I think a good way would be to allow both variants in a flexible manner. One could explicitly share his taxonomy with a number of users to make sure that his taxonomy always mirrors the taxonomies of the other workgroup members. A company could then install a central server for taxonomies and file storage and such. This would enable the users to share their taxonomies with this user (the server would exist as a user in the network...) and be sure to work with the same taxonomy as everybody else.

… I am drunk … its probably all rubbish :-D

Other Interests

I am personally interested in using problem solving algorithms within knowledge based systems. That includes using ontologies to model the knowledge and have a reasoner to perform logic based resolutions or decision trees or neural networks or <insert technique here> to find similarities and to roughly classify unknown data.

To find the way back to SnipSnap: such a system could help an expert to classify his document into the existing topic schema. It could roughly analyse the text, search for keywords, and learn "online" how to classify the documents. It would then propose the classification to the expert. That would ease the classification a little and lower the barrier to classifiy at all.

Sense or Non-Sense?

Give me more!

Thinking about PModes idea: Having a "give me more" button would be quite useful indeed. The system could select all documents, which are classified in a similar way and return those. A question which remains open is: What does "similar" mean? Maybe as a kind of specialization? For example: one snip is classified as "Java" and another snip is classified as "Java Applets". A user who visits the "Java" snip and wants to learn more would be presented with a kind of navigation panel showing the next more specific snips - including the one classified as "Java Applets".