Parc Experiments on WikiDashboard, Tagging and Search @ KMWorld

November 9, 2007

in Knowledge Innovation

Ed Chi and Lawrence Lee are presenting on Web2.0 in the Enterprise. They are from Parc see this page for relevant research links. . They want to focus on more near term  things you can do today. I captured this during their Wednesday session and only realized today I hadn’t posted it.

They have noticed three key trends.
1) web as a platform for services
2) rich UI technologies like Ajax and Flash
3) Activities together generate Collective Intelligence (this third trend is likely to cause the most disruptive impact.)

What they like about web 2.0 tools!
Knowledge is captured and archived in the course of administering the information and in the course of having conversations online. He’s referring to wikis blogs, bookmarking etc. However, some of the costs are forgotten. eg interference costs, interaction costs (cause friction, and noise in folksonomies. Their research has looked at consumer uses like del.icio.us. Then looking at cognitive models so we can recreate and then look through living laboratories online.

Augmented Social Cognition.
This is a focus of current research. What does it mean for cognition? To remember, think and reason, that can enhance how a group and make better decisions faster. They are working on three different projects that sharing.

WikiDashboard: To bring forth Social Transparency!
Declining percentage of activity devoted to article editing in Wikipedia Lot of talk about implementing them in the enterprise. So we ran data on Wikipedia, downloaded it all, created a database. Showing a stack graph. Over time they are seeing reduced time on articles, consistent talk around articles, a few more users, more user talk and huge increase in other (eg rules on which wikipedia works / procedural pages),  and maintenance. So non article interference is now up to 35%. New coordination costs; anytime someone reverts a page it is seen as a slap in the face!

So WikiDashboard is a social dashboard for wiki.  This is a plug-in for wiki installations that provides a visualization all the social patterns and dynamics that relates to that page. You can then drill down into activity history etc. This can reduce conflict and coordination issues. Showing an example of Hillary Clinton. Obviously a controversial page. Shows top editors and the overall editing activity over time. The percentages in relation to edit provides an interesting example of who’s really contributed. You can also do pivot browsing. It’s rather interesting to see their other interests and assess the writer’s motivations. It becomes even more important to look into the social dynamics and the individual actions within an organizational wiki. In the enterprise it is more important to make it visible into the activities.

SparTag.us
Interaction Costs and Social Systems: You should see the work of  author “The wealth of networks” Interaction costs will determine the number of people that will participate. By lowering the transactions costs you will naturally up the participation. When you think about distilling that idea down to a single graph as you lower the cost of participation you increase the number of people that will create something for you for free.  In the enterprise people want to spend even less time. In this setting it is important to keep the interaction costs very low. One of the problems with Delicious… he says he stopped using it. It is too costly. I go to a webpage and then type in some key words etc. It takes too long. While very successful as a lightweight system we think there is a way to innovate on the UI.

I can’t find a working or downloadable example of Spartag.us on teh netw. The demo shows a possible method for reducing the friction in the current process of tagging. Example on techcruch and reading an article on facebook. The idea is that i want to just click on the words and then just highlight and tag on this page. As soon as I tag . The idea is you just click and highlight what you have already been used to doing in knowledge work. The idea is to make the social tagging activity as lightweight as possible and make it part of people interaction of reading. Click to tag, and Paragraph tagging. the technology works at the paragraph level. You can also see what your co-workers are reading etc. They are also incorporating tag probability. So if you read a newstory it could be on more than one website. So anytime it says.. here is the same para somewhere else it is tagged to the content rather than the destination.

A problem in folksonomies is they are generating a lot of noise. By letting users build the tag it makes discovery easier. Example of how tags aren’t consistent. eg – how to’s, tutorials, turorials,  these inconsistencies and noise create many problems in delicious type systems. YEP including my own.

You should think about tagging as a communication. With tags they are encoding topics as tags so users are then applying them as navigation markers. Unfortunately the tag in the middle may also generate noise. They found that tags contain less information about documents over time. The ability of a tagging system to deal with this problem gets worse and worse. (Snowden said the same and it’s true).  Example java. in the beginning you will find specific documents about java, as it becomes more and more used it will get worse and worse over time. Some words will become useless navigation markers.

Just like in search engines. User have more and more tags per bookmark. So more and more keywords per search is happening. As words become less powerful they coping by adding more tags. So how do you deal with these noise problems. This sets up their third example

TagSearch:
When you start a search it goes thorugh all the bookmarks and then returns the common tags that this word is related to. So we can provide an overview. Then selecting the ones that look most interesting and likely for java /ajax and instantly the system responds with some relevance feedback; implemented on top of this tag search engine. They are using semantic analysis to build up the results. Conclusions: In order to effectively deploy it is will require real research in the interaction and coordination costs. You should look at increasing the transparency in the organization. In spartagus you see a system that reduces the tag cost and then in serach a system to reduce the amonut of noise.

These last too were interesting demonstrations. I can’t find them on the web I don’t think they are ready for prime time yet.
 

Tags: , , , , ,

Previous post:

Next post: