Hello Alexander, --- "Alexander \"SquidLord\" Williams" <[EMAIL PROTECTED]> wrote:
> Pietro: > > I read your proposal with great interest, but I > don't think the idea > of tagging with an approximated power set/cloud > really matches the > expectation of how people actually tag. I know its > not how I tag, if > it matters. Are you sure? I mean, you use many tags and for sure some are mor important than others. If we approximate a list with a power law, and we define the steepness to be a function of the lenght of the list, then the power law is bound not to be very steep for the small list we are used too. Also on LJ you use the memory function. I use to have a lj too. AT the time, on lj you were allowed to only use 5 terms. For the rest they were similar to tags. If you need to use only one term what will you use? (put that in the first place) Only two terms what will you use (add those two as the second and third terms). And so on. It is actually very intuitive: use all the terms that are relevant and (if possible) use them in order of importance. If thay are not in the correct order it's not the end of the world. Those tools are designed to be flexible. But I try to be more specific later. > > Essentially, the difficulty is the issue of "order" > in terms of my > conception. If I tag a post as "fiction, prose, > science," I'm not > trying to say that the fact its prose fiction is > more important than > the fact it's about science. On the contrary, I > CAN'T say, because > that meaning is ultimately in the mind of the > community as a whole, > not in my perception. On the other hand we know that for each URL delicious returns us with a tagcloud which is just a position in an n-dimensional hypercube. Imagine only three terms, and 5 URLs as 5 points inside this cube. Delicious provides you with the info of where the url is. Which than let you do neat things, like fine neighboring URLs. I think there is nothing wrong in giving to users the possibility to point to which area of the cube they see the URL as belonging. Maybe there are better ways to do that. Maybe it would be better to introduce also parenthesis: so (fiction, prose), science means fiction and prose are equally important. And the interesting thing is that people could choose to insert this information or not. Some people would not care to insert the tags in an order, and delicious would still work fine. Some people would insert the order to specify the rough position in the hipercube, and they could use metric among their own URLs, and delicious would also work fine And finally some people could use the order and the parenthesis and the position in the hipercube would be even more precise. The way in which it is coded is not really important, I suggested to use the order because it seemed to me sensible, easy, ignorable at the beginning, and would give a rough idea of where in the hypercube a URL would be. What instead I consider important is: give to the users the possibility to specify the position more precisely. > By assuming my tag set is really a cloud, you're > injecting meaning where there really isn't any. > Further, if Del > started assuming tags are input as clouds rather > than sets, it would > actually limit my expressibility. I can't say, in > such a context, that > an entity is equally two or more (or potentially a > lot more) things; > there's an implicit hierarchy. > > I thought we were trying to get away from > preconceived hierarchies? my view: We are trying to build a tool that works. This at the moments seem to suggest that we should avoid preconceived hierarchies. As a way, not as an aim. I don't see why letting people specify where in the n-dimensional hipercube the URL lives would induce a hierarchy. (beside the point that it would be an 'induced hierarchy for each user' and not a 'preconceived hierarchy equal for everybody'). Even if you decide to tag every 'python' snippet of code as 'python programming' and every Cpp as 'Cpp programming' you could still tag a paper that discuss programming, in particular python as 'programming python'. It is the lack of considering the order that, by collapsing the space of possibilities, induces a hierarchy. Not the reverse. And in fact del induces a hierarchy, and you know it, since you are using the mindmap maker which has been designed using that intuition. > > >From my perspective, the tag cloud as an entity can > ONLY occur post > hoc, once the community has commented on the nature > of the entity with > tag sets. Assuming (or even requiring) that the > attached notation is a > cloud rather than a set invalidates a lot of the > meaning you can > derive from the resulting hypercloud. Hmm, why? The sum of tag clouds is a tag cloud. It might be a different tag cloud, mapping the same URL to a different position, but I don't see why it would hold less information, or less relevant one. > It means LESS > that, knowing the > annotations are clouds, that the word that comes up > most often first > is "prose," rather than in the current state of tag > sets, the word > "prose" occurs most frequently. You are saying that this would mean that 20 people tagging something as 'prose' on the 20th position would be less important than one single person tagging it as 'prose' on the first and position. That might be true, but the 20 person will have other tags on the first position, which will (remember they are all renormalised) lower the importance of 'prose'. So we are really discussing about fifth position decimals: a difference of 0.000001 in the tagweight. Not really a big deal, plus it has not been proven that the new tag cloud that gets generated as a sum of tag cloud instead of a sum of tag sets wouldn't be better. And as a final point, no one will stop you to just completely ignore the order position and still get exactly the tag clouds as we are having now. You could even study the difference of the two type of tag clouds. > > This isn't intended as a heavy criticism of the idea > of tag clouds as > meaningful. On the contrary, I think the post hoc > observation of tag > clouds over time is THE best means of understanding > how the community > thinks of an entity. I disagree that its worth the > cognitive overhead > to think of tagging with tag clouds rather than > sets. > Thanks Alexander, I hope I answered your points one by one. Pietro http://blog.pietrosperoni.it/tag/del.icio.us __________________________________ Discover Yahoo! Get on-the-go sports scores, stock quotes, news and more. Check it out! http://discover.yahoo.com/mobile.html _______________________________________________ discuss mailing list discuss@del.icio.us http://lists.del.icio.us/cgi-bin/mailman/listinfo/discuss
