Sorting web pages by inscribed tags is never reliable: authors can lie, may be wrong, do mistakes, and mostly they do not tag pages. Results of such semantic searches leave most of good stuff tailed and also it can easily head you with some irrelevant pages, and the best of all, it can leave your query empty, in case where non-semantic service would give you the best with absolutely same query.
If we create web service where we can tag our found material as we identify it. We may built user interest groups, where we are more sure the others had searched similar stuff and tagged it in similar semantic sphere, having tagging skills at "our level".
This points out two teoretical things:
1) Tagging as giving some level of semantic access to different materials over the web is process which is familiar to language: every tagger has different tagging experiences, notions and needs, and finally every tagger would tag random file of materials using different principles. We can also be sure, that any tagger will easily find taggers having same qualities, lets say same dialect, if he will have access to mechanical (find similar) and also social methods (theme chat, news, forums) to get together with similar taggers.
2) Tagging is more flexible when for one document we have N tag files, and more effective, when for our tag files are externally grouped, what gives us chance to select organic set from possible tag ratings, and also select and sort results after computing proportion of this set.
1) Tagging as giving some level of semantic access to different materials over the web is process which is familiar to language: every tagger has different tagging experiences, notions and needs, and finally every tagger would tag random file of materials using different principles. We can also be sure, that any tagger will easily find taggers having same qualities, lets say same dialect, if he will have access to mechanical (find similar) and also social methods (theme chat, news, forums) to get together with similar taggers.
2) Tagging is more flexible when for one document we have N tag files, and more effective, when for our tag files are externally grouped, what gives us chance to select organic set from possible tag ratings, and also select and sort results after computing proportion of this set.
There is not much problems to solve this, and also, what is pretty, it can go such way, so that most of internet users would use it, also along with google, yahoo and similar services. How would that looked like? Take it fast: many webservices around the world would enable users to tag any page or link on their own, some of them could also be connected to browser (technically possible over few saved columns of data per service and some service access standards not too much!), so you could click you browser to tag page, or it could continue in today's "share it!" spirit, what is, being honest, crap. So imagine you need to find fan club of irfanview where plugins are free and listed to download. You will open your service search window, log in, select user group "Best freeware geeks, iv forever", than write down your semantic search and with some probability you will find it. If not, you will open google, will dig it out, and click to your browser "Tag it" button. You will select service, group, create tag description, and select if you want to publish your prey as group news or only at your service public tlog.
Inscribed to Tim Berners-Lee
and open or free creatives around The Web.
*People are always telling you what to do, but what's right for them may not be right for you.