I’m a participant in a number of different blogging and online communication forums and it’s very interesting to compare their evolution as new features of blogging move into the forefront. I’ve been particularly watching the development of tagging in the last year or so too, with recent articles about de.licio.us and Technorati, among others.
That’s why I was so intrigued by the implications of the recent message I received from my colleague Tim Ribich of IT Toolbox, wherein he detailed the following policy change for this thriving IT blogging community…
Dear bloggers,
Starting this Wednesday, Dec 7th, your Moveable Type blog publishing tool will require you to input keywords for new posts before setting them to “Publish” status.
Keywords (similar to subjects or categories) aid in getting your posts listed in blog aggregators and searches under relevant topics. Effective use of keyword tagging will help you connect with targeted new readers who share your professional interests.
After completing each new article, consider what words your target reader might use to search for the information you’re posting, then input those separated by commas within the keywords section. Please limit yourself to 5-7 of the most highly relevant keywords for each post.
This certainly makes sense to me: using a keyword mechanism helps categorize articles and makes synchronistic discovery easier too (and it’s also a smart way to make the article just a wee bit more search engine friendly too, with yet another occurrence of the key concepts in the article).
In an interesting parallel, Chris Pirillo and the team at Lockergnome are also requiring the use of keywords, though not for Technorati. Instead, Chris helped create a “tiny footprint” search engine system called Gada.Be that the Lockergnome blogging system automatically uses for its keyword presentation.
Are we seeing the rise of tagged keyword taxonomies (or “folksonomies”) after all, even though I earlier predicted earlier that Tags are terribly implemented in the current Web?
Hmmm…. I’m still not convinced. What do you think?
Users should be able to gradually build up their own taxonomy/folksonomy. Basically it would be a semi-structured and prioritized list of key phrases that matter to the user. Every text editor that wants “keywords” should take the user’s list of phrases as an input and then text mine the keyed text and present the user with a list of the key phrases that occured in the raw text and then the user can decide whether to accept all of them, edit them, take the highest priority n, etc.
One feature to support is synonyms: if I use a phrase, it means a specified tag. For example, the text “post to a blog” might imply the tag “blogging”.
Users could “seed” their personal taxonomies by borrowing a friend’s or selecting one or more base taxonomies from web sites.
The user’s taxonomy would be kept as a “taxonomy stack” so that the user will automatically pick up new tags as they are added to the master taxonomies.
A weak-AI automatic text miner could also scan text provided by the user (e.g, papers or articles you wrote or your email folders) and produce a semi-structured list of prospective key phrases that the user might opt to add to their personal taxonomy. The auto miner might also be driven my a master “mega” taxonomy build from the universe of taxonomies to guide it in selecting phrases from the user’s raw text.
The concept of “clustering” could also be used to “suggest” a prioritized list of possible categories for a block of text based on the user’s personal taxonomy. The user might then pick one or more of the suggested categories and that would refine the filter priorities for the recommended keywords for the text.
The process of managing the user’s personal taxonomy is a bit more complex than that, but the process of “tagging” a block of text such as a blog post *must* be as simple as I’ve suggested.
Manually retyping keywords is simply for the birds, and no *courteous* user interface would require any decent user to do so.
I’m personally all in favor of tagging, taxonomies, folksonomies, etc., but the emphasis should be on automation such as mining rather than requiring tedious and error-prone manual tagging by the poor user.
Let me be clear: Computers are here to do as much work as possible *for* us, not for *us* to do work for them.
Out of curiosity, how come you didn’t “tag” this post with the keyword “folksonomy”? My automated scheme would *never* have missed that one.
— Jack Krupansky
Technorati Tag: “Manual Tagging Considered Harmful”
Dave – I’ve always believed “tagging” as most people know it, has a very low ceiling and low probability of success especially in large information sets. Tagging with an architecture that leverages the idea of typed associations, has a much higher ceiling and greater probability of success. MyST Topic Cloud demonstrates why. An example: http://blogsite.com/topics
Jack – I like automation as well, but there are cases where automation may not be suitable. Amazon specifically understands why humans are good at certain things and should be factored into the equation – http://mturk.com.
bf
Dave and Jack, of course you are both right, tags (currently) are implemented terribly. But, the fundamental need to map information spaces for findability is age-old and will not go away.
See:
A Holy Grail: Effective Knowledge Capture, Persistence, Enhancement, and Transfer
http://myst-technology.com/mysmartchannels/public/item/53612
Current tagging schemes are, by and large, extremely weak, but they are a step in the right direction. There are *many* ways to improve on current tagging schemes, but perhaps the most important is to introduce the notions of “typed taxonomies”–or more accurately, “ontologies”–and typed associations.
The current wave of “tagging” is really just a very simplistic form of “topic mapping”, but without any notion of taxonomies, ontologies, or typed associations. The field of topic maps is well established and is even backed by several international standards such as ISO/IEC 13250.
See:
The Topic Map Standard
http://www.infoloom.com/tmstands.htm
MyST co-founder Bill French and I actually met in a former life (at Starbase Corporation) where together we ran a research project (code named “Elmer) focused on applying topic map technologies to information systems.
See:
Elmer Preview
A Standards-Based Architecture for Creating a Unified Information Space
http://www.idealliance.org/papers/xml2001papers/slides/Seidl/Elmer%20Preview%20-%20XML%202001%20Conference%20-%20Orlando.ppt
Bill and I left Starbase to found MyST Technology Partners with a vision of creating a platform that would, one day, make it possible to apply some of these principals to mainstream information systems. We’re pretty excited about the fact that just two weeks ago, we released the beta version of what we call “Topic Clouds” that take the idea of “tagging” another baby step closer toward the topic mapping ideals.
See:
MyST Topic Cloud� Released to Beta
http://blogsite.com/public/item/108414
To learn more about Topic Clouds, you can find a briefing paper here:
http://myst-technology.com/documents/papers/MyST%20Topic%20Cloud–Product%20Briefing.pdf
MT is attempting to solve their problem of bad URL generation for posts.
Their assigned URLs may not be considered Search Engine Friendly.
So they may forecast that having bloggers enter keywords into their posts, this will boost the SE results rank, and the visibility of the MT software in the blogosphere.
Are guidelines given re: spamdexing?
To answer your question: in my judgment, no, taxonomies and tagging are marginally relevent, for remedying poor IA or nav tools.
If your blog contains rich relevant rare content, posts that powerfully flow from passion, expertise, and writing skill, you needn’t worry about SE optimization or tags or folksonomies.
If you have well written, search engine friendly URLs, you should be fine in your esoteric niche.
My sites usually come up on top of almost all vital keywords, with no effort at SEO or any other social networking system.
Content + SE Friendly URL = High Rank
Dave, back in my Library Science days I studied the advantages and disadvantages (both real and theoretical) of controlled vocabulary versus uncontrolled vocabulary based indexing and retrieval systems. From an efficiency and precision basis “controlled” usually won out. But the ultimate professorial comment always seemed to be “index in the language you expect your users to use.”
While these principles may not have changed, the difference now is that retrieval systems can, behind the scenes, quickly and efficiently map and remap the users’ language to a variety of taxonomies, all in real time. The result seems to be that controlled vocabularies (as implemented by those seeking to control tagging vocabularies) appear to be less important than they might have been.
I’m not buying totally the resurgence of uncontrolled vocabulary indexing, though. The existence of multiple indexing schemas, word mappings, classification systems, and taxonomies means that you are now dependent on the semantic technology being used for retrieval to determine how “good” or “bad” your particular retrieval instance deals with “fuzzy language.” My guess is that different taxonomies will continue to evolve and that, even if services such as Google implement easy to use tagging taxonomies, authors and users will continue to evolve. This is, after all, the Internet, not a nicely controlled internal system. You can never be sure, no matter what controlled indexing schema you use, that it will be identical to the language being used bby the user.
Meanwhile, I continue to use my own blog system’s tagging feature, which means I am building multiple sets of tags (called “categories” by Squarespace). I’d rather not have to worry too much about that side of things too much since I’d rather spend time creating clearly written content. I’m still guessing that this is the best determinant of whether I will be “found” online.
– Dennis D. McDonald (http://ddmcd.squarespace.com/)
For an example of tagging taken to an extreme in a playful and, one might even begin to imagine, useful way, check out what the guys at Robot Coop have been up to:
http://www.43things.com
http://www.43people.com
http://www.43places.com
allconsuming.com
My hesitations of using tags are related to the multiple content created within the blog.
As this tagging process produces many similar pages that repeat the same content over and over again.