Folksonomies 2.0 - The Chaotic Order
Categories: Information Architecture
Tags: folksonomies, folksonomy, ia, social_classification, tag, tagging, tags
I have a tags related idea that periodically comes back to my mind. Yesterday I had a chat with Peter Van Dijck about it.
Folksonomies are a very widespread concept today and also a few big magazines have understood their revolutionary approach and value.
What I’m asking to myself is: “Since a year ago, which evolution has emerged? Which new ideas are people working on to improve and empower social distributed classification?”
Before explaining which new features I’m thinking of, I would like to explain why do we need an evolution of tags at all.
I love tags. I really love, as an user, having a way to add a simple powerful metadata layer to my data without having to adopt a centralized hierarchical schema proposed by someone else.
But tags, as we now experience them, don’t satisfy me! They are messy and their widespread adoption is making them also more messy.
Take a look to popular tags’ page on Del.icio.us. How far is this from a hierarchical structure or a nice faceted strategy? Maybe too much. Too much because it’s not easy to extract a mental model out of this stuff and if you cannot use your tags, they will soon loose their power.
Yes we could add an arbitrary structure and semantic to create a better and more powerful UI over these tag clouds, but the rule #1 about folksonomies is: “leave your users free“. Your users act on a personal basis but the aggregation of their mental associations is like a live organism. The structure emerges bottom up by an implicit, quite magical, consensus.
Are you sure that is not possibile to have a better browsing interface without imposing an arbitrary (so limited, restricting, not scalable, biased…) structure?
I’m not sure so I’m here to propose a different more powerful yet not restricting approach.
Faceted Tags
In its wonderful Mefeedia, Peter has introduced the concept of Faceted Tags. They are great! Among other nice improvements, Peter has identified a few basic facets (Place, Topic, Language, Event, People) and every tag is associated to one or more of these facets.
The result is that you can surf tags using facets. Really powerful. We can access the big mess through a new order!
These facets could remind us of Ranganathan’s PMEST super facets. They are general in a specific system. What I mean is that choosing the right set of facets at the beginning let us descrive completely a corpus of items.
But what if I’m only interested in all videos from “Italy” or “Rome”? I’m happy because Rome is one of the tags provided under the “Place” facet, but if I really want all videos from Italy? or South of Italy. I’m still facing the big mess. I have to search for all tags and see if I can find some labels that remind me of Italy. Or suppose that I’m searching for a monument in Rome without knowing its name. Europe or Italy is not enough, I would like to have two tags “Rome” and “monument” that I can use together!
Tags Hierarchies
We can go a step further: people choose tags, Peter chooses facets (now at least..). Now imagine people also creating a hierarchy of tags under a facet: I tag a video with “Rome”, but Rome is a city in “Italy” and Italy is a country in “Europe”. I have a way to create a hierarchical order between tags and facets become real polyhierarchies that I can intersect to reach a better set of information. Why bothering to create this structure? Simply because this structure gives us a context to better understand the semantic meaning of the tag and a better tool to browse the tag world moving from specific to more general kewords without loosing the ability to reproduce users’ mental model.
Wiki Wiki Tags
Now again a step more. Take the city of Venice. Venice is in Italy or in US? Both! So who decides where this tag has to be? Simple again: the users! Venice is a city in Italy and is a city in Utah. We are in the digital word and an item could be in different places at the same time. In a wiki based approach, users could iteratively move tags from a part of the hierarchy to other parts or create “copies” of the same tag in different places. If you do a search for the tag “Venice”, you will get two different results. If you browse the tags’ hierarchy, the presence of two tags with the same label is probably transparent for you. So a hierarchy is emerging again through a bottom up consensus.
All together now!
So I’m talking about a new folksonomical approach in which tags are assigned to (and searchable by) facets. Every facet is a hierarchy of tags emerging from the actions of the users.
Users can search or browse tags. More facets and more tags can be used together at the same time following an iterative process of successive refinements.
What do you think about it?
In the same category:
February 18th, 2006 at 1:31 pm
Folksonomy 2.0
Ultimamente sto leggendo una marea di articoli sul sistema delle tag. C’è un grande interesse da parte della comunità sull’argomento. Il sistema di classificazione dei contenuti presenti sulla rete attraverso le parole chiave sembra esse…
February 19th, 2006 at 12:01 pm
The hierarchies-of-tags idea reminds me of the Dewey Decimal System that also allows the combining of tags (330 for economics + 94 for Europe = 330.94 European economy) but in a more flexible way.
Do you think users would be the ones who compose the hierarchy, or would a computer be better at generating this? Or experts?
And can you explain your idea of “user ould iteratively move tags from a part of the hierarchy to other parts”? That sounds to me like a recipe for disaster. The alternative, copying tags, sounds better, although some kind of “see also” system of references would be required.
I haven’t looked at the basic facets that Peter van Dijck uses, but “Topic” to me sounds like the most general facet ever…
In closing, I like the general idea, but it needs some more thinking, and maybe some examples? If you’re really into the “2.0″ spirit, a quick beta would be the best thing
February 19th, 2006 at 1:36 pm
Thank you Peter for your answer.
As I said in the post, the tag structure (and its variations) should emerge with the users’ action. Forget experts. Probagly the system could support this process too. In my opinion this is not a recipe for disaster but instead the opportunity to let an organism-like organization to naturally emerge.
For the facet point: Topic is only one of the facets (the one associate to the Matter category in Ranganathan). Other facets applies here (the instances of the Ranganathan’s categories time, space, personality, energy, etc).
As for the quick beta.. Well it wouldn’t be so quick, at least for me to code it! I will try to convince Peter Van Dijck to experiment these approach..
See you on Friday!
February 22nd, 2006 at 9:56 pm
Faceted Tags
I love the idea of giving tags dimensionality and depth. Emanual Quintarelli writes about faceted tags and Folksonomies 2.0. Somebody stop me before I fall down the rabbit hole. I just started tracking some links, starting in Emanual’s blog entry…
February 23rd, 2006 at 7:42 pm
Great article Emanuele, thanks for writing it. I think hierarchical categories created by users could help with the findability issue, however, I wonder what that would do to the “discoverability” characteristic of tagging and folksonomies. Usability may also be an issue until tagging becomes more integrated into the browser/os, etc.. I would be curious to see what the level of interest would be for this among general users at this point.
February 27th, 2006 at 10:25 pm
Hierachical tags are interessting, but i don’t know if people are able enough to that.
And if the hierachy is from the users, maybe it becomes too complex. Namly everyone have is own hierachy. In your example “Rome”-”Italy”-”Europe”, maybe you want to add the province. I, a Dutchman, don’t know the province, so I do’nt want that in the hierachy.
February 28th, 2006 at 9:26 am
Maarten, the hierarchy is completely optional and the UI to add the hierarchy must be really thoughless for users.
I will add a few details to this post from the talk I gave to the first Italian IA Summit later today. Anyway:
1. The hierarchy doesn’t need to be shared between all the users (you could have a personal hierarchy as it is now with common tagging)
2. I could add a province, but you are not obliged to use it: Rome will always be under Italy. So If your level of granularity (related to what you know on the matter) is lower, no problem! You will continue to use the part of hierarchy that is interesting and useful for you.
February 28th, 2006 at 9:39 am
Ellyssa,
the hierarchy introduction would have the following benefits:
1 - Improve the findability (hierarchies add context and help clearing the tags’ meaning)
2 - Improve the serendipity and the navigability (after you have checked all the content described through the tag “Rome”, you see that Rome is under Italy and you can start to review also the content tagged with Italy and so on..)
About the usability concern: here the UI design is fundamental. We have to study this problem deeper, trying to come out with a good solution..
What do you think?
March 2nd, 2006 at 11:20 am
Emanuele, this is a stimulating post.
It is interesting to see that people feel in need for some merging bewteen
traditional (hierarchies, facets) and new (folksonomies) tools. Maybe there should be some software managing the links between tags and a structured scheme. The usual way of doing this is allowing people to search for a subject through a box, and bring them to the point in scheme where it occurs, so that they can navigate starting from them. But this is no more a folksonomy, it’s a thesaurus/taxonomy
The problem with not having a central authority is how to keep the structure consistent enough to be useful, e.g. using the province level or not.
Also: are people willing to care about specifying facets and hierarchies, not only quick tags?
(I agree with Peter B that Topics means everyting apart from the extra-subject facets (like Language or Target), and should not be an equivalent of Matter, rather it should include PMEST.)
March 6th, 2006 at 11:01 pm
I think that with critical mass and finely tuned feedback mechanisms social bookmarking systems will indeed be able to build tag hiearchies on the fly based on a tag weighting algorithm. I have noticed that with high volume bookmarks on del.icio.us if you start at the bottom of a Common Tags listing and work your way upwards you are going from the more specific (less consesus) to the more general (high consensus) tags. Therefore you could build a hierarchy in reverse order.
March 7th, 2006 at 7:31 pm
James,
I agree with you that if you read a tag cloud in reverse order you start from more specific and go to more general tags.
What I cannot understand is how you can come out with a hierarchy only through tags’ specificity.
Weight can be used as a clustering principle but I’m not seeing how to leverage on it to produce a father-child structure.
March 15th, 2006 at 10:27 pm
Folksonomies 2.0
One of the best speech heard yesterdey during the IA summit was the one of Emanuele Quintarelli about Social distributed classification. He addressed the problem of a good classification in the information overload (and chaos) era, starting by the emer…
March 16th, 2006 at 8:07 pm
I think the problem with any effort to clean up tags is that users simply won’t use them. Tags are best used by users when they are a quick hit. I’m reading this article about Rome and I just want to tag it Rome, so I can remember it later. Forcing users to go through Italy then Rome will increase the likelihood of the user saying screw it.
While open tagging may be ugly, I think its the only way that will work from a user entry point of view. If categories and heirarchies are established then they can only be done by the system (which would be ugly at best).
March 16th, 2006 at 8:15 pm
Dave, I’m not sure of having catched your though completely but my intent wasn’t in the direction of a tags’ normalization.
My Idea is still leaving users free of tagging exactly in the same way they are doing now. You can still add the tag ‘Rome’ to this post without creating or knowing at all about the existence of a tag ‘Italy” that contains the tag ‘Rome’.
But If a third user wants to add put Rome under Italy, he can do it. The system at this time will aggregate this information, eventually exposing a hierarchy to the community of users.
And also consider, that one of the main points here is the user experience.
Take it simple, but not too simple.
Does it make more sense now?
June 14th, 2006 at 9:05 am
On the behalf of Paolo Massa:
I read “Automated Tag Clustering: Improving search and exploration in
the tag space”
( http://www.rawsugar.com/www2006/20.pdf )
that says:
“With RawSugar, taggers can specify tag hierarchies in
their own accounts (saying that sushi is a subtag of food for
example). The system uses these hierarchies to provide a
strong exploration and search experience. Figure 3 is the
RawSugar tag box extracted from a search page for restau-
rants on a user’s account. The tags are grouped according
to the user defined hierarchy and thus present a more pow-
erful exploration space.”
You might want to explore rawsugar.com
or read the paper.
June 14th, 2006 at 9:15 am
Paolo, thank you for your comment.
I’m talking every day with people from RawSugar and I love their tool.
Frank Smadja (one of the authors of the paper you cited and the VP in RawSugar) has organized the collaborative tagging workshop at the WWW2006 this year.
I invite you to have a look also to other papers presented there: http://www.rawsugar.com/www2006/taggingworkshopschedule.html and to my post http://www.infospaces.it/wordpress/topics/information-architecture/91 about “Hierarchical taxonomies from flat tag spaces”.
Emanuele
August 14th, 2007 at 5:11 am
It is interesting to see that people feel in need for some merging bewteen
traditional (hierarchies, facets) and new (folksonomies) tools. Maybe there should be some software managing the links between tags and a structured scheme.
November 1st, 2007 at 9:18 am
Hi Emanuele,
I could not find your email address, so I’ll abuse a blog comment entry for this
I find facetag very interesting. I have written my masterthesis on Image Retrieval, and thus about tagging (and the problems that come along with it) versus content-based image retrieval. I figured you might be interested in the thesis? You can download it at http://photoindex.thingsdesigner.com/pdf/photoindex_mathesis_matthijsrouw.pdf
Tell me what you think ..
November 2nd, 2007 at 2:55 pm
“Folksonomies 2.0 - The Chaotic Order” - Good work. Cogratulations
February 11th, 2008 at 10:03 pm
How to Use Computer Data Recovery Software
Often our computer files are messed up because our computers will write our files when we are done using them wherever there is available space. This is why it is important to use computer data recovery software. Computer data recovery software will ma…
April 2nd, 2008 at 7:21 pm
Did you know there is a new cool and intimate new sushi place in Rome which offers high quality Japanese foods for eating or take-away, and offers great hand-made cakes and free wifi to all customers? http://naoko-sushi-roma.blogspot.com/ Click to Visit