The Structure of Collaborative Tagging Systems

a study of tags and the patterns of use.

協調タギングシステムとしてのdel.icio.usの解析。

利用者とタグの関連や，ウェブページのブックマーク数および付与された個々のタグ数について解析。

Model for collaborative tagging

Regularities in user activity

"tagging is like filtering"

Levels of description vary (e.g.: computing > programming > javascript)

Uses delicious data.

Explores relationships between #bookmarks, avg. #tags (unrelated)

Studies popularity of a bookmark over time, most of the bookmarks arrive quickly: "Of the 212 popular URLs in our dataset, 67% reached their peak in popularity in their first 10 days, 17% on the first day. However, at the end of the spectrum, 17% were in the delicious system at least 6 months before reaching their peak of popularity"

Consensus emerges after some time on which are the tags that an item should have ... the tag set for a particular URLs stabilizes. This is not explained entirely by imitation.

HP's research on del.icio.us

One of eariest work analyzed tagging activity in social tagging systems which includes some important insights.

interesting introduction

_EGC/colab/tags.pdf

http://127.0.0.1:800/_myd/exec.php?str=_EGC/colab/tags.pdf

Good list of tag functions:

• Who/what about
• What it is
• Who owns it
• Refining categories
• Qualities or characteristics
• Self-reference
• Task organizing

Other insights:

• Tag functions range from useful-to-others to useful-to-self.
• Levels of expertise and need affects tagging
• Tagging proportion tends to settle down
• Number of tags not related to number of bookmarks
• Popularity externally-driven
• More general tags used first

A good article with two distinct parts. 1) A discussion of some of the categorisation issues with folksonomies in the introduction, 2) An analysis of how tags are used on the del.icio.us bookmarking service.