Monday, September 24, 2007

Finding relevant resources and users using FolkRank ordering

Another major extension contained in the recent BibSonomy release is a
new mechanism to retrieve relevant resources and users for a given tag,
called "FolkRank".

From your previous experience with BibSonomy and other collaborative tagging systems like del.icio.us or flickr you will know that a typical way of browsing is to select one or more tags to retrieve a list of resources (e.g. bookmarks) ordered by the date of their addition to the system. This is very useful to discover the most recent material; but in some cases, the goal is to retrieve the most relevant resources for a given tag.

Consider as an example the tag "www". Obviously very relevant bookmarks for this tag are e.g. the WWW conference websites (e.g. http://www2006.org/, http://www2007.org/, ...). This conference takes place every year and brings together researchers and business people from all over the world to discuss trends and developments in the area of the World Wide Web. Though being highly relevant, these bookmark do not necessesarily appear in let's say in the top-10-list of bookmarks tagged with "www" (http://www.bibsonomy.org/tag/www).

For finding (amongst others) highly relevant resources for a given tag, our research group has developed an algorithm called FolkRank (see http://www.bibsonomy.org/bibtex/24d8b4f79814691fbe6db8357d63206a1/stumme).
Its idea is similar to Google's PageRank algorithm, i.e. it analyzes iteratively the link structure between users, tags and resources in order to calculate relevancy.

You have now access to this great feature via the URL http://www.bibsonomy.org/tag/www?order=folkrank or by clicking "order by folkrank" on a tag site. When you have a look at the result (sticking to the example of the tag "www"), you can see that highly relevant bookmarks like the WWW conference websites are found at the very top of the retrieved list. Furthermore, FolkRank is able to calculate highly "relevant" users for a given tag:



This information can give you a clue who might be an expert in a certain research area, and might thus support you in networking with colleagues interested in similar topics as you are. Have fun and success with this new tool!

Best,
Dominik

Popular Posts