Brainstorming
From Faviki Wiki
Contents |
[edit] Subjects
What kind of objects do you tag ?
- a whole website (example: google.com)
- an article (example: renault.com/products/megane.html)
- a file (example: megafileupload.com/65465/mypic.jpg)
- a bookmarklet (example: URL behind the Faviki button)
- a quicksearch (example: http://www.google.com/search?q=%s)
- an action (example: the web service that starts my coffee machine)
[edit] Predicates
By now, a Faviki item links a particular URL to a few DBpedia concepts, using the "associatedTag" predicate. Other predicates could be more expressive. Feel free to add any imaginable predicate.
[edit] isA
The subject can be a search engine, an online database, a blog, etc. This predicate is more expressive than using a tag. If I tag an URL with a "search engine" tag, how will others know whether the URL is about search engines, or whether it IS a search engine ?
waliitech.com isA resource:Job_search_engine
[edit] targets
For whom the web site operates. It is interesting to know whether a website targets your country or Luxemburg or Taiwan.
africa-jobs.com targets resource:Africa
[edit] isRunBy
gmail.com isRunBy resource:Google
[edit] represents
waliitech.com represents resource:Waliitech_Inc.
[edit] listOf
In case an URL does not represent a thing but multiple things.
list-of-conferences.net listOf resource:Conference
[edit] isExampleOf
Another relationship I have some trouble expressing, still thinking...
jamendo.com isExampleOf Web_2.0
[edit] follow
This predicate has no object, and just indicates that I (the URL poster) need to check it later. A "private" predicate could be imagined in the same way.
[edit] Objects
- DBpedia concepts (implemented in Faviki)
- me (I am a quite important concept in my life, and I could be identified by a Faviki URL)
- folksonomy tags (wild expressiveness)
[edit] Wikipedia subsections
Someday we could include the finer-grained subcategories for each Wikipedia article.
(I think Wikipedia errs in offering so few links to offsite resources, so Faviki is accumulating these and should eventually be merged into Wikipedia itself. At that time they will need to be broken down by subsections (and content types, below). When I add links to Faviki now, I'm targeting that future use, and limiting my entries to resources that Wikipedia readers generally should find useful. --RobotWisdom 08:54, 9 September 2008 (UTC) )
[edit] Content types
'Subjects' (above) describes filetype or media type, but most tagged objects will be articles, which can be further subdivided into categories like:
- essay, opinion, criticism, literary criticism, comparison
- announcement, alarm, preview, tip
- expose, gossip, anecdote, hypothesis, theory
- prediction, conjecture, suggestion, discovery, innovation
- history, biography, profile, eulogy
- anomaly, counterevidence
- accusation, satire
- interview, transcript
- tutorial
- short story, poem
- photostream
- map, painting, cartoon, graph
These are orthogonal to the Wikipedia categories but I think we need to systematise them and include one or more as (faceted) tags.
[edit] Identifying every concept in the universe
Wikipedia's aim is not to have a separate article for every concept in the universe (reference: [1]), so I am wondering: is there an effort out there to identify every concept in the universe ? How could such an effort be started ? As a layer over Wikipedia or OpenCyc ? Nicolas_Raoul 03:04, 13 October 2008 (UTC)
- We plan to add concepts from other vocabularies - Freebase and Crunchbase, namely. However, identifying every concept in the universe is a great challenge. I guess one way would be letting all the webpages on the web represent concepts, and allowing community to choose the best candidates. Vuknje
- Connecting the LOD (Linking Open Data) would be extremely cool, but I guess the user interface right now is very well suited to Wikipedia, and it will be a lot of work to make it as user-friendly for others sets. Also, I sometimes feel the need to attach one (or more) person identity to an URL, and FOAF identifying URLs would be the obvious way to do this. Nicolas_Raoul 13:22, 15 October 2008 (UTC)
[edit] Changing semantics
When a Wikipedia article is renamed, either the new name has never been used in Wikipedia, or it had been used in Wikipedia but is no more in use. The first case is just a technical matter, I am not too worried. What I want to talk about is the second case.
We have a problem when the semantic of a title changes. For instance, "Laden" used to refer to a fridge company, but since September 11 it refers to the guy who got famous. We need to differentiate between potentially several semantic meaning of a single title. If 10 years ago I tagged an URL with the fridge company tag, it probably has nothing to do with the guy.
I have been thinking about this and realized that it applies to anything using Wikipedia URLs, not just Faviki. We need a "proxy" server, let's call it proxy.wikipedia.org . To identify the dog concept, Faviki would not use wikipedia.org/Dog but proxy.wikipedia.org/Dog/n where n is a positive integer, initially zero for all article. When an article gets renamed to a name that never existed before, then no problem, its identifier is proxy.wikipedia.org/<name>/0. When an article gets renamed to a name that existed before (and that's where we have a real problem) then the identifier is proxy.wikipedia.org/<name>/1
It is generally not a good idea to have an identifier contain anything about the semantic behind, so the identifier could also be just a number attributed to an article. The "proxy" would provide a way to go from the identifier to the article on Wikipedia.
Nicolas_Raoul 13:59, 15 October 2008 (UTC)
[edit] Different ways to express the world
I have found Interwiki drifting, an interesting tool to explore semantic gaps between communities that have different ways to express the world. Useful work to estimate the feasibility of unifying several languages' Faviki data. Nicolas_Raoul 08:20, 17 April 2009 (UTC)
