in the second session of the tagging workshop ...

Emerging Motivations for Tagging: Expression, Performance and Activism

Alla Zollers

http://www2007.org/workshops/paper_55.pdf

• explores emerging motivations for tagging in the public sphere -- with tags that are funny, political, commentary
• requires insider knowledge to decode language
• opinion/expression, performance, activism

e.g defectiveByDesign tag on amazon.com

three kinds of tags before -- but they don't really overlap [this study didn't look at del.icio.us

tagging as a way to review or express opinion -- maybe a star system is enough? -- but not as creative?

jt -- again contextual knowledge again is important to define meaning

scott: evidence that others are using long strings -- tags copied from others

Applying Collaborative Tagging to E-Learning

Scott Bateman, Christopher Brooks, Gord McCalla and Peter Brusilovsky

http://www2007.org/workshops/paper_56.pdf

Scott and Chris present

looking at e-learning as an environment; student-generated tags; applications

• 'learning-objects' in e-learning
• "tags provide insight into learner's comprehension" -- John and Seligmann 2006
• look at tags to try and understand what learners do and what they are comprehending
• experiment in tagging based on pages created for non-specialist AI students
• compared student tags and content-author tags (specialist and non-specialist language)

results

• 50% of author tags are not within the student tags.
• only 25% of author tags were in the head of student tags
• text mining had a higher-level of agreement with students 68%

jt -- why are these two voices being presented as conflicting in an either/or scenario?

tagging vs annotating

• tagging most often at resource level; annotation at a finer grain (linked to a part of a resource)
• demo of annotation system (plug in to any learning management system)
• demo of iHelp tagging of multimedia - tagging video at a particular time-slice -- c.f. mojiti.com as another example

discussion

retrieval vs resource description goals -- highly influence tags supplied

Learning User Profiles from Tagging Data and Leveraging them for Personal(ized) Information Access

Elke Michlmayr, Steve Cayzer and Paul Shabajee

http://www2007.org/workshops/paper_29.pdf

presented by Elke Michlmayer

• folksonomies provide fine-grained information about users' interests
• show change over time
• used del.icio.us data (tags and dates)
• tag frequency as simple approach, but combination and co-occurrence revelatory
• weighted links to reveal strong and weak links (time critical to impute this)
• co-occurrence compared to co-occurrence with a decay factor
• metric based on Kendall Tau (compare members of two ordered sets)
• decay factor adapts to user's activity pattern
• tool to visualize changes in profile over time
• user / tagger mis-match (shows tags and keywords don't correspond)
• shows tags clustered related to events in users' lives
• issues: quality is user-defined

thomas: what happens with co-occurence when one term drops out of a common cluster? could this show changing semantics and vocabulary evolution

problem of two semantic spaces: user terms vs keyword terms (from the source --hp tech reports in her study)

Towards federated Web2.0 sites: the FolkDesk approach

jon iturrioz, Oscar Diaz and Cristobal Arellano

http://www2007.org/workshops/paper_34.pdf

Problem: user has lost global view of own resources, when tag in different application spaces.

TAGMAS = tag management system

• queries remote services and integrates results

questions:

how to expose the integrated results to the web (rather than only on the desktop)

tagging within the operating system -- also could be a means of leverage

why RDF -- tagging services don't give you enough data to build a robust semantic network?

jt -- doesn't it all depend on where you think your user space is? i could use a federated web identity -- but only if it let me selectively link -- or is that reveal -- the parts of my persona i wanted to....