Tagxster relies on a very small set of rules. Really. The basics are these :

1. Sentences have multiple meanings. Not only that but beyond their intrinsic meaning, they can signify many different things according to who read them.

2. Tags are not topics, they're tags. As such, they can be used in many a different way which is part of their undeniable appeal : nobody knows exactly how one is going to employ them. Are they proxies for actions, dates, themes (memes ?), topics, names ?...
   
   Ex : 43Things is a website whose peculiarity is to make extensive use of action-oriented tags.
   
   

   
   

3. One word can have various meanings. Thus, to put things simply, a tag is expressed by one (or many) word(s) which may mean different things. Accordingly, users should thence be granted the possibility to tag more than one thing with different tokens of the same word corresponding to two different tags.
   
   Ex : I got misled once by the word "rare" which means two very different things in English : "cooked just a short time" and "uncommon". Just because a passage from a culinary recipe was tagged with the word "rare" should by no means prevent someone from using the same word, with its other meaning, as a tag. Of course, the word "seldom" is still available but this is rather fortunate, if not fortuitous ; furthermore, what about users who have got very good reasons to use "rare" instead of "seldom" ? Eventuallly, whatever those reasons might be, a tool, powerful enough, should allow for the possibility to distinguish between the many senses of a word.

4. Different contexts induce different meanings. Who would deny that heterogeneous contexts have the capacity to produce at least slight variations ? Then, of two words with the same meaning, one will end up having a different meaning than its counterpart as soon as it is made to appear in a different context, especially one that's specifically designed to induce such variations.

5. Whence the importance of tagclouds provided they're no longer only restricted to their customary usage as a visualizing device. Tagclouds, or rather, "tagspaces" as we shall now call them, are better suited as context inducers, even more so since it brings a significant shift between previously "passive" tagclouds operated by blind algorithms, to user-driven meaning spaces. We shall delve more into that soon.

Where does that lead us ? An examination of Tagxster's features.

Tag'n Highlight.

First, to the possibility of tagging part of a text with whatever fitting tag thanks to a system of multiple "tag-layer" highlights.

Ex :

All that is needed is to select a tag and highlight those parts you see fit. Click the tag again and you'll visualize highlighted marks corresponding to the areas falling under the selected tag. Choose another tag and this time all you'll catch sight of are those areas corresponding to the new tag.

Stacked layers warrants that different (and potentially conflicting) colors are no longer needed as surrogates for tags. Thenceforth, a single passage can be highlighted as many times as needed, hence accounting for the fact that a text won't necessarily appeal the same way to two different persons or evoke the same responses, depending either on its intrinsic plurality of meanings or the readers' own personal interests and idiosyncratic backgrounds.

Ex : Let's have a look at this highlighted text.



Judging from it's meaning, we may decide to tag it with the words "namespaces" or "folksonomies" for example. Now, on a more personal note, since my thesis dwells on similar topics, why not tagging the same passage precisely with the word "thesis" (in an attempt to gather all the interesting bits I come across under one tag, for back-up reasons for instance) ? This is something highly idiosyncratic because a) not everyone is a PhD candidate, b) even if you are, our theses subjects are more likely to differ than the contrary. Rather than merely "already there", this kind of meaning is over imposed and extrinsic (therefore, it is also something no natural-language retrieval algorithm could ever infer).

Bookmarking

With a great many "Web 2.0" bookmarking websites, people never get inside the things they're indexing or sharing. All in all, documents are tagged in an "extrinsic" fashion. On del.icio.us, for example, even if you happen to carefully peruse every item behind a link before adding tags, this preliminary step isn't integral to the tagging process itself while, with Tagxster, those very passages you decide to tag'n highlight will automatically lead to your indexing the document you're exploring (the tags weight being pondered according to how much text a single tag subsumes). One action doesn't simply lead to another one : both are the same under two different guises.

DaTagbases

Now, Tagxster doesn't limit itself to the exploration of only one text, nor to indexing a collection of discrete documents devoid of any connections amongst them. Using one or more tags while working on two texts will inevitably introduce new relationships between them.

Especially since what the tag'n highlight feature leaves open, rather than merely limiting ourselves to indexing links, bibliographic references, or whatever one-dimensional object pleases you, is a cross-indexing of those highlighted bits. With this system, each tag indeed de facto compiles all the passages (we can't talk of mere "citations" any longer) a user sees fit to gather under it, not regarding their miscellaneous origins. This feature, namely the capacity to stock all the bits originating from different sources and formats under a great number of tags is tantamount to creating as many personal databases.

Being based on the same pool of documents and reflecting one's multifarious interests, these "tagbases" may also, amongst other things, make it possible to benefit from the comparison between different passages related to a single tag (remember that this tag may embody all manner of things : topics, actions, headers, etc.)

Tagspaces

Let's finish this presentation with tagclouds. Tagclouds can now be found almost everywhere on the Web. At best, they're a useful and easy to the eye alternative way to envision lists of items. Of course, there are some interesting variations like Newzingo, but the fundamentals are still there : users don't have control over tagclouds, the latters being reduced to a mere passive display of information. But things could be different. Wholly different. Considering the gap between how items are stored and how they're accessed, tagclouds do not affect the former. Which is fortunate as it lets us create as many tagclouds as we want. What for will you ask ? Well, I don't know. Or rather, I know exactly how I would like to use that feature.

However, how precisely you may find that useful remains unknown to me. But the rule is there and anyone may contrive useful usages of it. Instead of trying to cover every possibility by devising beforehand an exhaustive lists of categories or options, let's empower users and present them with as many possibilities as the rules governing them are simple. So, what are the specific rules for tagspaces ? All in all, there are three of them :

1. You may create as many tagspaces you wish and put inside just those tags you want to work with. For instance, all the tags representing topics and the likes might be bundled together in one tagspace while the tags referring to actions and/or dates are brought together in another one, as in faceted classifications (see below).
   
   
   
2. However, it is possible to drastically change these groupings at any time for tagspaces are not folders, contrary to the latter the former allow for adding or deleting tags with no constrains whatsoever.
   
   
3. Tagspaces may be given a title to induce a new context of signification : thus, two tags represented by the same word, used the same way, may be slightly affected and, consequently, diverge from one another.
   The principal governing this rule amounts to turning facet classifications on their heads. Some sites already make use of both folksonomies and faceted classifications. Mefeedia's faceted tagclouds are a very good example of this.
   
   
   
   The bottom-line is this : all the tags in either the "Places" tagcloud or the "Languages" have been put thither owing to some intrinsic characteristic they retain. That's from all the tags it contains (which are limited to towns, countries, etc), that the "Places" tagcloud derives its name.
   
   Now, contrast this with tagspaces. Since tagspaces are only empty places to begin with, they can no longer get their names from preexisting tags or categories only. Imagine an English Scholar, let's say a specialist of Henry James (whose concept of "figure" played an integral role in his famous short story The Figure in the Carpet). As such, following the master's steps, this scholar, in this example of ours, would specifically wish to be able to discern "themes" from "motives", the two words being invested with very different meanings for the sake of his studies (what developer could design specific categories to cover such peculiar cases ? ).
   
   To make sure this is possible presupposes that two instances or two tokens of one and the same word may be used as a theme and as a motive. After all, why should a writer fear the limitations that are now prevailing in this so-called Web 2.0 of ours ? "Good" or "Goodness" for example, might appear in a novel both as a theme and a motive, with no ensuing conflict arising. Still, the fact that one tag belongs to one tagspace or the other, in other words, that it has a specific context, obviously affects its signification. This time, meanings are induced from the outside rather than the inside, from a context entirely set out from the beginning by users.
   
   
   
   As we've previously seen, there are two conditions to realize that : first, one needs to be able to use one and the same term in two different contexts. While it's difficult to imagine two faceted tagclouds with exactly the same tags and different headers, it's not a problem with tagspaces. Then it is also required that the number of tagspaces should not be restricted so as to offer users enhanced freedom to devise conceptual tools adjusted to their needs.
   
   Tagxster will ( ;-) ) do just that (provided you want it to, of course).