Delicious hacks CSS Fig 3. Rules with two elements of K2 with 0.05 support, and 10% confidence described in Section 3. Due to space restrictions, we have to focus on a subset of projections. In particular, we will address the two projections Ki, witI 1:= id and a2:=(1→1,2→3,3→2). We obtain the two dyadic contexts K1:=(U×R,T,1)with1:={(u,r),t)(u,t,r)∈Y}andK2:=(T×U,R,I2) with I2: =I(t, u),r)l(u,t, rEYI An association rule A- B in KI is read as Users assigning the tags from A to some resources often also assign the tags from B to them. This type of rules may be used in a recommender system. If a user assigns all tags from A then the system suggests him to add also those from B Figure 2 shows all rules with one element in the premise and one element in the conclusion that we derived from K, with a minimum support of 0.05 and a minimum confidence of 50%. In the diagram one can see that our interpretation of rules in KI holds for these examples: users tagging some webpage with debian are likely to tag it with linur also, and pages about bands are probably also concerned with music. These results can be used in a recommender system aiding the user in choosing the tags which are most helpful in retrieving the resource later Another view on these rules is to see them as subsumption relations, so that the rule mining can be used to learn a taxonomic structure. If many resources tagged with rslt are also tagged with rml, this indicates, for example, that rml can be considered a supertopic of slt if one wants to automatically populate the relation Figure 2 also shows two pairs of tags which occur together very frequently without any distinct direction in the rule: open source occurs as a phrase most of the time, while the other pair consists of two tags (ukque and ukg: irc), which seem to be added automatically to any resource that mentioned in a particular chat channel The second example are association rules A B in K? which are read as Users labelling the resources in A with some tags often also assign these tags to the resources in B. In essence both resources have to have something in common. Figure 3 shows parts of the resulting graph for applying associationhttp://pchere.blogspot.com/2005/02/absolutely-delicious-complete-tool.html http://www.onlamp.com/pub/a/onlamp/2005/01/20/rails.html http://www.onlamp.com/pub/a/onlamp/2005/03/03/rails.html http://rails.homelinux.org/ http://www.onlamp.com/pub/a/onlamp/2005/06/09/rails_ajax.html http://www.rubyonrails.org/ http://www.rubyonrails.com/ http://www.cssbeauty.com/ http://www.cssimport.com/ http://www.webstandardsawards.com/ http://www.thenoodleincident.com/tutorials/box_lesson/boxes.html http://pro.html.it/esempio/nifty/ http://home.tampabay.rr.com/bmerkey/cheatsheet.htm http://www.alistapart.com/ http://www.scifihifi.com/cocoalicious/ http://tuxtina.de/software/ http://www.inknoise.com/experimental/layoutomatic.php http://www.accessify.com/tools-and-wizards/list-o-matic/list-o-matic.asp http://www.evolt.org/article/Ten_CSS_tricks_you_may_not_know/17/60369/ http://www.positioniseverything.net/ http://www.citeulike.org/ http://www.connotea.org/ http://www.fiftyfoureleven.com/resources/programming/xmlhttprequest/examples http://openrico.org/home.page http://www.baekdal.com/articles/usability/usable-XMLHttpRequest/ http://script.aculo.us/ http://www.adaptivepath.com/publications/essays/archives/000385.php http://www.ajaxmatters.com/ http://developer.apple.com/internet/webcontent/xmlhttpreq.html http://www.modernmethod.com/sajax/ http://www.webpasties.com/xmlHttpRequest/ http://www.xml.com/pub/a/2005/02/09/xml-http-request.html http://jibbering.com/2002/4/httprequest.html http://jpspan.sourceforge.net/wiki/doku.php?id=javascript:xmlhttprequest http://goog-ajaxslt.sourceforge.net/ http://www.ripcord.co.nz/behaviour/ http://prototype.conio.net/ http://johnvey.com/features/deliciousdirector/ http://tool-man.org/examples/edit-in-place.html http://tool-man.org/dragsort/ http://www.netlobo.com/div_hiding.html http://tool-man.org/examples/sorting.html http://www.bobbyvandersluis.com/articles/goodpractices.php http://www.formassembly.com/ http://www.axentric.com/posts/default/7 http://brothercake.com/site/resources/scripts/dbx/ http://www.ajaxpatterns.org/index.php?title=Main_Page http://www.onlinetools.org/articles/unobtrusivejavascript/index.html http://www.onlinetools.org/articles/unobtrusivejavascript/ http://leftjustified.net/site-in-an-hour/ http://toolkit.crispen.org/index.php?cat=templates http://www.stunicholls.myby.co.uk/index.html http://supergreg.hopto.org/nutritious/nutritious.php http://delicious.mozdev.org/ http://dietrich.ganx4.com/foxylicious/ http://www.beelerspace.com/index.php?p=890 http://kevan.org/extispicious http://opencontent.org/oishii/ http://fresh.homeunix.net/delicious.html http://pchere.blogspot.com/2005/03/great-flickr-tools-collection.html http://glish.com/css/ http://www.airtightinteractive.com/projects/related_tag_browser/app/ http://www.gamingheadlines.co.uk/wod/formstyle/index.html http://www.kryogenix.org/code/browser/sorttable/ http://www.quirksmode.org/ Ajax Delicious Hacks CSS Javascript Fig. 3. Rules with two elements of K2 with 0.05 % support, and 10 % confidence described in Section 3. Due to space restrictions, we have to focus on a subset of projections. In particular, we will address the two projections Kσi, G with σ1 := id and σ2 := (1 7→ 1, 2 7→ 3, 3 7→ 2). We obtain the two dyadic contexts K1 := (U×R, T, I1) with I1 := {((u, r), t)|(u, t, r) ∈ Y } and K2 := (T ×U, R, I2) with I2 := {(t, u), r)|(u, t, r) ∈ Y }. An association rule A → B in K1 is read as Users assigning the tags from A to some resources often also assign the tags from B to them. This type of rules may be used in a recommender system. If a user assigns all tags from A then the system suggests him to add also those from B. Figure 2 shows all rules with one element in the premise and one element in the conclusion that we derived from K1 with a minimum support of 0.05 % and a minimum confidence of 50 %. In the diagram one can see that our interpretation of rules in K1 holds for these examples: users tagging some webpage with debian are likely to tag it with linux also, and pages about bands are probably also concerned with music. These results can be used in a recommender system, aiding the user in choosing the tags which are most helpful in retrieving the resource later. Another view on these rules is to see them as subsumption relations, so that the rule mining can be used to learn a taxonomic structure. If many resources tagged with xslt are also tagged with xml, this indicates, for example, that xml can be considered a supertopic of xslt if one wants to automatically populate the ≺ relation. Figure 2 also shows two pairs of tags which occur together very frequently without any distinct direction in the rule: open source occurs as a phrase most of the time, while the other pair consists of two tags (ukquake and ukq:irc), which seem to be added automatically to any resource that is mentioned in a particular chat channel. The second example are association rules A → B in K2 which are read as Users labelling the resources in A with some tags often also assign these tags to the resources in B. In essence both resources have to have something in common. Figure 3 shows parts of the resulting graph for applying association 7