Help talk:Relation

Relations have a lot of overlap with the current category system. All articles with a specific relation could be stored in a cached list so it doesn't have to use any more computational resources than the current category system. Are categories needed at all in a semantic wiki? Any thoughts? Fuelbottle 00:56, 15 November 2005 (CET)

Well, categories have one nice built-in thing: There are not categories of categories, which makes them very OWL-compatible or in other words: Categories are categories and not just a page with a funny name. An article can be in a category but a category cannot. So categories are really categories of artices and not something that groups something.--140.203.154.12 01:43, 15 November 2005 (CET)


 * Categories can indeed contain other articles AND other categories AND even themselve. Semantic Markup can also used as category system, but it's mutch more than just this. While the category system describes just categories, semantics can describe also the kind of the relations within and between the categories and articles. MovGP0 02:28, 15 November 2005 (CET)


 * I guess Fuelbottle is right. A category system can replaced by semantic markup, but we need special coding, so that a link  is handeled equal to  . The most important thing for the beginning will be to be able to handle both in an equal way, so that we can choose a soft migration from the current system. Later we could even let Bots replace the old system when a user uses it and replaces step-by-step the current system - but at end the users decide, and I gess many of them will want to keep the current system because there are familar with it.
 * The caching is theoretically easily done (even the implementation won't). All the semantic Information will be stored in an special Database. There will be at least a seperate 3-colums Table containing semantic Triplets. Dependend on the Implementation we could separate special Attributes/Properties, as well as derived ones, into separate Tables. Ie. after a sematic Reasoner has analyzed and simplified the relations between them. How this will get handeled depends on the Programmers - hopefully they will wisely do so.
 * Note: I'm not a member of the Programming team. If you want to figure out the Details I recomend to learn at least RDF, PHP, and SQL - and reading the Sourcecode.
 * MovGP0 02:47, 15 November 2005 (CET)

In fact, we will re-use existing Semantic Web technology, so the database design is no longer our problem. The equivalent is called a Triple Store, a specialised database for RDF data.--Max Völkel 03:53, 15 November 2005 (CET)

unparseable Values
I recommend to show all strings that are not parseable just as strings (so the user can interpret it) and mark them in a graphical way as unparseable (red color, a star, etc.) and offer a Link to a Page witch describes what is able and what is on the whishlist. MovGP0 12:31, 15 November 2005 (CET)

Types I want to have at minimum: MovGP0 12:38, 15 November 2005 (CET)
 * Strings (beginning and ending with ")
 * Integers, Floats
 * Boolean (True, False)
 * Dates (31/12/-1917; 31/12/1819; 1770)
 * SI units

How to do this?
Example: (Please excuse the last statement - it's wrong, but it makes the example tricky)

Bill Clinton was President of USA from 1993 till 2001. George Walker Bush was President of USA from 2001 till 2009. Bill Clinton was President of UN from 2009 till 2012.

Question: how can we model this relations within the Wiki Syntax?

I'm thinking about a syntax like these: meta:= (was president of:=USA),   (was president in:=1993),   (was president in:=2001) meta:= (was president of:=UN),   (was president in:=2009),   (was president in:=2012) But this seems not to be very elegant to me - has anybody a better idea? Thanks, MovGP0 18:01, 15 November 2005 (CET)


 * Perhaps by letting relations have attributes, for example by letting attributes of a relation be specified by adding :attribue to the relation? E.g if you have the relation "is a President of USA". This relation could then have a year attribute with range in parantheses specified by a start and stop value, i.e. for Bill Clinton

is a::President of USA:year:=(1993,2001) is a::President of UN:year:=(2009,2012) If attributes could have sub attributes, then perhaps the year attribute could have an optional subattribute to and an optional subattribute from, and the syntax might then be something like: is a::President of USA: year.from:=1993, year.to:=2001 is a::President of UN: year.from:=2009, year.to:=2012

Fuelbottle 22:16, 20 November 2005 (CET)

I think the whole example below doesn't make much sense. It's soo complicated. Current semantic web technologies cannot deal with time and space in a general way. Our goal in Semantic Wikipedia was to encode the trivial, obvious factual knowledge, and make that searchable. Okay, the natural language query: 'Who was president of the USA in 1999' doesn't sound very far fetched. But to explain to a machine the exact semantics of the example below is really hard. Who was president of the unite states on 5.7.2002 ? We humans read the table quite sophisticated. We know, there is always onlyone president, we know presidents work for more than a year. If we hadn't that knowledge, we could answer: 5.7.2002 ? We don't know. We know only who was president in 1993 among others. So this Bill was one of the presidents. Really, these things are tricky. Why not leave them out for a while? --Max Völkel 18:53, 22 November 2005 (CET)


 * Sure that's tricky - therefore I'd this Post to show it. But I have already stated a possible solution how to model this: Talk:Attribute:from:

{from date:=20/1/1993}{to date:=20/1/2001}{was president of::USA}
 * witch maps to:

 20/1/1993 20/1/2001  
 * Search is a bit more complicated. To figure out who was the President of the USA in 2001 you could make the following query:


 * witch returns both, Bill Clinton (was president) and George W. Bush (is president).
 * Even such things might be complicated to implement, it could be a very powerful solution.
 * MovGP0 21:09, 22 November 2005 (CET)

has type vs. has Type
Hi, i've just experienced a little problem and wanted to point someone on this to see if it's a bug or a feature ;) I had problems creating types for attributes because i copied the text from this page. After trying a little i figured capitalisation IS important. I think the page should say: has type::Type:Geographic area

instead of

has Type::Type:Geographic area

It's a very little difference (no capitalisation of "type"), but after changing it i didn't get the no-type-assigned-to-attribute error anymore. I think there are also other occurences throughout the page. Hope this helps someone ;) Besides this, great project. Trying to keep up with all the stuff you implemented.