[Tagdb] Single and multi-word tags / swarming and spreading

[Nitin Borwankar]

Timothy Spalding wrote:

 > I submit that this shows two things. (1) Naive users expect to be  
able to use multi-word tags. (2) Experienced users (particularly  
experienced users with a background in information theory) > may  
gravitate to single-word tags.

Hi Timothy, Juan,

Yes, a very lively debate indeed.   Could I add a possible nuance to the 
2 points above?

The way I read Timothy's earlier email is as follows :-

It's not just that LibraryThing's users may be naive users, but that 
they may actually prefer wordy descriptions being of a literary bent.
So its possible that the linguistic preferences of the user,  rather 
than the "maturity" may be factors.  This begs an explanation as to why 
more technical users use single word tags.
My explanation is that more technical users gravitate towards / prefer 
terse, abstract descriptions rather than wordy ones because of their 
technical background and quickly evolve to a single-word-tag usage. 
Whereas, I hypothesize, even more mature users of a literary bent may 
continue to use more descriptive language best expressed in multiword 
tags.   For them expressiveness rather than linguistic efficiency is 
paramount and my theory is that they may never change that even after 
significant "experience".

As implementers we need to be sensitive to the different linguistic 
styles of our user communities and do what's appropriate for them.
Forcing everyone into a "terse mode" doesn't seem like the best way .....

Nitin

> Juan,
>
> Great, a debate! That was a fun email to read. Your linguistic point  
> was very welcome and well taken. (I did Classics in graduate school,  
> so I relish moving in a linguistic direction.)
>
> I feel, however, that you're talking about what ought to be, not what  
> is. You write:
>
> "At the time, it made a lot of sense that multi-word labels should be  
> supported, but the more I tag, especially in del.icio.us, the less I  
> care about having terms I see as bound together stored as a unit."
>
> I submit that this shows two things. (1) Naive users expect to be  
> able to use multi-word tags. (2) Experienced users (particularly  
> experienced users with a background in information theory) may  
> gravitate to single-word tags.
>
> It seems to that EVEN if I concede the superiority of single-word  
> tags to multi-word tags AND pass over the problems of "Tiera Del  
> Fuego"—something I'm not inclined to do—that users' actual behavior  
> wins the point. The central principles of web design should be "don't  
> make me think" (ie., a "low cognitive load) and "do what I want it to  
> do." All these go against imposing any "system" on a user.
>
> Put another way: in my head, that article is about "early America."  
> Who are you to tell me it's about "early" and "America," or force me  
> to call it "earlyamerica." Butt out already and let me tag my stuff!  
> I thought this website was supposed to be easy! Forget this. Let's  
> see who won American Idol last night.
>
> Tim
>
> On Apr 10, 2006, at 11:17 AM, Juan Cristian Vera wrote:
>
>> Hi all;
>>
>> My name is Juan Cristian & I've been lurking for a couple of  months. 
>> I thought this would be a good time to jump in:
>>
>> I designed a social schema for work last November. Our main goal in  
>> making a schema from scratch was to be able to do some  collaboration 
>> that exposes a sort of atomic trust-based  transparency model, 
>> allowing you & chosen communities in which you  participate to tag 
>> your stuff on our site (internal stuff, you  construct within the 
>> site - I call it a "flamazon" model - the  marketing guys don't think 
>> that's a good word...), and see other  tags based on the three-way 
>> trust relationship between the owner of  the tagged object, the 
>> tagger, & the viewer...
>>
>> This behavior would look a little like "swarming", except that you  
>> can set the reslution of the tag cloud you see while you navigate -  
>> whether you are tryinhg to reach somethign based on your tags  alone, 
>> those of one or more communities you belong to, or the coud  at large.
>>
>> I started reading the tagdb list when I first launched into this  
>> project, not knowing much about tagging as a user. Since November,  
>> however, my browsing habits, and the way I look at classification  in 
>> the first place, have been profoundly affected by practice in  
>> tagging and some of the thought-provoking discussions on this list.  
>> To wit, the first one that I ran into was a post about commas and  
>> spaces in tags. At the time, it made a lot of sense that multi-word  
>> labels should be supported, but the more I tag, especially in  
>> del.icio.us, the less I care about having terms I see as bound  
>> together stored as a unit.
>>
>> I respectfully disagree with Tim. I think the relatedness of two  
>> words that one comes to see as one term is a form of clustering,  
>> almost the beginnings of rhetorical structure. I mean, if I see a  
>> picture of a dog house and label it with these two tags, dog &  
>> house, I will be quite able to discern when I go looking for  
>> something what are pictures of dogs in peole houses as opposed to  
>> dog houses. The same problem occurs in spoken language, not just  
>> with adjacent terms, but as well with homonyms. The fact that one  
>> term is subordinate to another, like a sattleite paragraph in  
>> rhetorical structure theory, does not make the two terms need to be  
>> bound more tightly tha by coincidence in the single tagged object.  
>> In fact, in omance languages they are not. "dog house" may seem  like 
>> a single word in english, but this composite noun, in Spanish  would 
>> be "la casa del perro", or without a contraction, "la casa de  el 
>> perro" with an article and a preposition that are implicit in  
>> English, possibly implying a closer relation between "dog" &  "house" 
>> than there actually is.
>>
>> Best,
>>
>>
>> Juan Cristián
>
>
> _______________________________________________
> Tagdb mailing list
> Tagdb at lists.tagschema.com
> http://lists.tagschema.com/mailman/listinfo/tagdb