Taxonomies and Tags

Thanks Joshua. Your posts are eerily relevant to a large project I’ve been working on for the last few months (launching in early August *knock on wood*).

This project is a community website based around an annual event and we have decided to create an official taxonomy or categories (used to classify and organize content and enhance search) as well as a folksonomy/tags system where users can tag content with any keywords that they choose. Those tags are only relevant to the user similar to the del.icio.us style of bookmarking and classifying content. At some point it would be interesting to be able to implement a trigger that tracks common user-generated tags and incorporates them into our official taxonomy.

Once we get this beast up and running we’ll have more time to experiment.

So an example of what you’re proposing is:

Say I have a detail page of an iPod.

And people start tagging it with “Apple, MP3 Player, Music, etc etc.”. Then say over a 1000 people tag it Apple. Then the site should auto make a category called Apple.

Is that what you’re talking about?

Dennis…that’s basically it. I think the function that isn’t currently there is the threshold of creating new navigation choices based on temporal tagging…this will help to cover more social or news related items…as those don’t often appear in taxonomies.

Or put it another way, once a folksonomy term reaches a certain level of popularity, it gets shuffled sideways into the ‘official’ taxonomy? if it falls out of favour over a period of time, does it get demoted?

Like Scott, I’m working on defining a project which leverages our expert knowledge in the form of pre-defined tags we use internally whilst also opening to user tagging to see what new shapes and connections the information gets formed into. I like the idea of a trigger.

Have a look at http://iNeedSomebody2tag.com/welcome/en. There is a web experiment regarding to folksonomies and collaborative tagging systems.

Maybe it is of interest for you.

Regards,
Tobias Kowatsch

I’ve been saying this for a while, but none of the clients I work with have thusfar really “understood” the concept of a fluid folksonomy and user contributed ontologies. Ultimately, a social website’s navigation should be defined entirely by some sort of algorithm based on user activity, but it’s not easy to implement or get right. So you need room to iterate post-launch. Which doesn’t happen very often in web production environments: you lay out a plan and launch. Then you just bug fix.

In a slight twist to what you’re suggestion Intructables applied a Taxonomy, half their own design - half folksonomy based.

Intructables: “Users may now tag their Instructable with two of eight categories: Art, Craft, Food, Home, Life, Not Liable, Ride, and Tech. Food, Art, and Tech categories are basically self-explanatory.”
http://www.instructables.com/

Tim O’Reilly had this to say: “This is where folksonomy meets taxonomy: the taxonomy provides a language that allows the poster to bracket the entry with tags that locate it in a coordinate space that is uniquely meaningful and suited to the subject.”
http://radar.oreilly.com/archives/2007/04/creative_catego_1.html

Josh, thanks for the link. If the world of language were only this simple that this worked consistantly. The folkonomy is a killer resource, but it lacks structure, which it crucial to disambiguating terms. There are algorithmic ways of getting close to this end, but they are insanely processor intensive (think days or weeks to churn out this structure). Working from a simple flat taxonomy or faceted system structure can be enabled for a folksonomy to adhere to.

This approach can help augment tags to objects, but it is not great at finding objects by tags as Apple would surface thousands of results and they would need to be narrowed greatly to find what one is seeking.

There was an insanely brilliant tool, RawSugar, that married taxonomy and folksonomy to help derive disambiguation (take appleseed as a tag, to you mean Johnny Appleseed, appleseed as it relates to gardening/farming, cooking, or the anime movie. The folksonomy can help decipher this through co-occurrence of terms, but a smart interface and system is needed to do this. Fortunately the type of system that is needed to do this is something we have, it is a taxonomy. Using a taxonomy will save processor time, and human time through creating an efficient structure.

Recently I have been approached by a small number of companies who implemented social bookmarking tools to develop a folksonomy and found the folksonomy was in far more helpful than they had ever imagined and out paced their taxonomy based tools by leaps and bounds (mostly because they did not have time or resources to implement an exhaustive taxonomy (I have yet to find an organization that has an exhaustive and emergent taxonomy)). The organizations either let their taxonomist go or did not replace them when they left as they seemed to think they did not need them with the folksonomy running. All was well and good for a while, but as the folksonomy grew the ability to find specific items decreased (it still worked fantastically for people refinding information they had personally tagged). These companies asked, “what tools they would need to start clearing this up?” The answer a person who understands information structure for ease of finding, which is often a taxonomist, and a tool that can aid in information structure, which is often a taxonomy tool.

The folksonomy does many things that are difficult and very costly to do in taxonomies. But taxonomies do things that folksonomies are rather poor at doing. Both need each other.

Thanks for the pointer, Pauric. That’s a nice example of this middle ground.

Using categories in the way they are reminds me of Lakoff’s idea of basic level categories. These are categories that we seem to find across most cultures and peoples…categories such as dog, child, food, etc. (the ones of little dispute)

Thomas, the structure you refer to is the hierarchical nature…correct? That seems right, as having hierachical structure helps us make sense of large sets of informatin.

The classification aspect of taxonomies (and the politics of classification) is what hurts people, when their vocabulary doesn’t match.

What I’m advocating, then, is a taxonomy with no politics…which has the ability to change over time according to the people’s actions, not the politics of classification.

Josh, taxonomy does not have to be a hierarchy.

I think you are confusing incomplete taxonomies, which folksonomies can help to augment, with the strong value a taxonomy provides. It is a similar mistake to confuse del.icio.us as the be all and end all for folksonmies as it has a long way to go to really take full advantage of what is there (they are working hard at getting a solid foundation in place to better enable the next steps).

Language is political by its very nature. Going back to Aristotle and rhetoric and semantics through time (one of the best semanticists of his time, S. I. Hiyakawa, became U.S. Senator) and its impact on Politics and lowercase politics. Words have meaning and different understanding through context. That is part of the value of a folksonomy, which helps highlight these shadings.

Really good taxonomists understand the shadings and social differences of understanding of the meaning of words and work hard to incorporate that into a usable structure for finding information. Decent taxonomists lean strongly on the “expert” terms of understanding and do not fold in the shadings of understanding and context.

Language (I am speaking of just with in one language and not the complexity of cross-language and cross-cultural (at a high level), but more on a broad simple level) is very complex. Human experience shades peoples understanding and relationships to words.

One of the problems with building large scale and broad reaching taxonomies is they are really difficult to build. Part of the difficulty is capturing what people call things in their regular usage, which a folksonomy does really well. But, the remainder of the hard work is building structures that encompass the shadings and mapping these with synonyms and thesauri.

I completely agree with Thomas (vanderwal). Folksonomies are great, they’re dynamic and intuitive, rich and surprising. But for navigation and topical coverage, you need something more formal, reliable and defined. If nothing else, people hate it when categories change too often and they can’t use the same path they used before.

So as far as I’m concerned, taxonomies and folksonomies are complementary tools rather than competing.