prime tags or iffy overlapping tags

Now we're talking - Hugh coined "Prime tags" (love that one) as the common denominator when Toby asked for a "master list" to be used when describing anything with fewest possible tags.

And Jon joined in with a definition of "Prime tags" being "tags that have more than less commonly shared meaning".

Are we not a bit stuck in the "bottom-up" thinking pattern here? A bit stuck in "tree structures" going out from a trunk? Precision is all etc.

Here we need to be precise, stumble once and everything is messed up.

The "experiment" suggested using multiple tags when choosing, then even iffy keywords (tags) will be able to deliver precise results.

"Top-down" thinking as it were.

Note that "salad" and "diuretic" alone could do the trick (what would I know with my limited botanic knowledge :). And there could be other and different tags that could cover the Dandelion with precision.

Maybe we have to do with more than 33 tags to cover the whole earth population or all the books ever written. I think we can live with a few more.

Example:

• Tag an object with "pattern".
• Some will think "music", some "maths" and your grandmother may think "knitting".
• Add "cloth" and it narrows down.
• Add "tartan" and at least I see some Scottish patterns in my mind.
• Still not precise? Add "Macleod".
• Or instead of "Macleod" you add "The Highlanders" and "Main character". Both of these highly different tags could lead to the same - but none of them alone would lead to anything specific.
• It's the combination that makes the difference.

Classic 'knowledge building' is based on common denominators, building blocks - or tree structures! Precise stuff required, sequence allimportant.

Multiple tags is the opposite, building knowledge from diffuse and imprecise tags/keywords using the intersection actively. Sequence does not even matter.
The more precise tags the better of course, but as these things goes one or two or three more tags makes even very iffy tags useful.
And going from 12 to 14 tags makes no big difference in practice.

Now you would know what that plant could be used for as well! That's "knowledge". Delivered by tags defining it.

For the user facing "mac", "macintosh" and "apple" knowing that all could mean the same could avoid these at first, then try one at a time at the end if required. (That's why we did the 'experiment' - trying to find a simple interface that would be easy to use.)

Important:

The author, the one adding tags, should be generous in sprinkling tags - it does not matter. In fact the more the better.

It is the user, the one who seeks the object that will choose how many tags he needs to make a precise choice, a precise intersection. In above example, using only "The Highlander" and "Pattern" he may find out what Hugh should wear at weddings!

Comments

[Ric]

Yes - we are slightly biased toward the hierarchical thought pattern, but we're getting there. What I said on the last post has been raised by a couple of people, and indirectly by you here - even with 'tagsonomy' selecting a single wrong 'tag' will miss the target. For instance, if someone were looking for "me" (i.e. not me, Ric; but someone who matches my skills, characteristics etc) and put 'female' as one of the tags, it would automatically exclude me. So it may be more useful to stop short of something that is too precise, and just use enough tags to restrict the view to a usable list of candidates.

This also highlights the importance of 'describing' what you want to have found in a manner that makes it easier to find - a set of tags that are logically related, but not too precise. The more precision an author uses to describe the object, the more precise the searchers have to be to find his work.

BTW - do you find it interesting that Hugh gets more comments about your post than you do?

[sig]

Ric, prexisely, let the user use his logic and the next one his proper logic to find the same thing!
For that we need many and even iffy tags - tagged by the 'object-creator'...

The only thing left then is an easy way to find the object - therefore the 'experiment': Sequence not important, click, unclick, play around let the logic fly... one guy finding a 16 tags tagged object by three choices, another with five others and so forth. Folksonomies meet structured knowledge :-)

Yeh, is it not annoying that Hugh is so popular? Hehe, it must be that he's such a cool, dapper chap... and that he's got the knack for giving stuff a good top-spin, being a great and funny writer and in general having a huge following! That must be it!
(Do not mention his cartoons compared with my crude napkin dabble... I rest my case :)

[Alex]

Sig,

One question do you really think tags can replace hierarchies? Sure hierarchies have problems, especially when wide, deep or named by someone else.

But by projecting context (and by the way tags are just one facet of context) over a file system you can remove a whole heap of irrelevant information from a persons view of the file system, suddenly the hierarchy becomes useful again as a way of grouping and making sense of the information that matches. This is what I am incredibly excited about.

Let me know if you want to see my thoughtpaper on this topic.

Alex

[Ric]

When I commented earlier I was still at work (staying late again!), and didn't have a real good play - but I just went back and had another look, and actually worked out what I was supposed to do (so tell Junior that the interface can't be too bad)!

OK so we seem to agree that the 'object-creator' should have as many 'fuzzy' tags as practical, making it easier for the object to be found. Now left to itself that could mean a very long list of possible tags (down the left of the screen) on your CMS test site. THAT'S where the interface gets nice - because you can trim the list of tags to choose from by (eg) entering the first letter(s) in the text box (but I notice that you can't enter a list of tags in the text box - is that a possible next step?). This may also help Alex cut out the 'irrelevant' info - although if the 'object-creators' are good, will there be any irrelevant info? Maybe 'not-required-at-this-time' would be more like it.

[sig]

Ric, Junior will be glad to hear :-)

When I tried it first time I made 35 tags for 34 posts... easy to get lost on left side and not so good for demo purpose. One solution we thought about is to let all highlighted tags stay visible (and de-highligtable) while working the little 'search' box... that might work.

Good question - will an 'object-creator' ever add an irrelevant tag? In the eyes of the 'object-seeker', absolutely! Another mind, a different logic - of course some tags will look funny. But for the 'object-creator'? As long as he's not mischiveous I do not think anything is irrelevant - it's his call after all! It's his creation, it's his concept and thus his relevances.

That he will think some of his own tags will be funny in two weeks time? Absolutely :-)

But it should not matter! As long as there is an interface that enables easy jumping around, trying out different combinations it should be OK... I think... at this stage...

Junior is full blast with version2 (funny those geeks, just finished and they hate the code already...) with a really neat managment interface (could open that one up then), much cleaner php code (according to himself) and of course some new feature tests...

And maybe we will add RSS and trackback too? Who knows... maybe it'll be a full fledge blog system by then :-)

(And of course, we'll flip the mysql 'faking to be an OODB' to the Thingamy kernel with its rather neat OODB and other stuff. Thus the 'experiment' could turn into being the basis for interface-navigation, blogs, websites and of course a nifty mail client for the Thingamy!)

[Chuck Turnitsa]

Very interesting. One thing to think of in a taxonomy - with a data (or linguistic) taxonomy (vs. a natural biology taxonomy, of bugs or fishes), it is possible for an item (or class) to have several spots on the taxonomy (in essence, it can have more than one set of parents, and then generate more than one set of kids).

To use your diagram - the concept of dandelion can be a child of the parent nodes "weed", "salad", "wine", "scent", etc.

When we use tags from such an non-determinstically developed taxonomy, perhaps we should discuss the lineage of the particular tag we are thinking of. . .

[sig]

Chuck,

one way to add flexibility to the thinking is of course to tag tags... "weed" being tagged with "plant" - choose "plant" and all "weed" will appear even if they're not explicitly tagged with "plant".

I try not to use (hard!) the words lineage and parent-child as they tend to push me towards a tree-structure image even if it was not meant that way - I'd rather try to think 'relationships' (even iffy) beyond 'being born by' or strictly belonging to a familiy or branch.

And the ancient Greek gentlemen did after all 'define' knowledge and understanding as being the set of relationships between objects. And I think the value of the knowledge will increase if we can go beyond explicit relationships (like in parent-child) and include peripheral and even unlikely relationships.

That's why I like to use 'iffy-blob-interception' as the object-definer :)