From: Daniel S. <dsc...@in...> - 2007-04-26 23:01:55
|
It seems to me that some of the variance in views here regards what each one understands as a "wiki", in this context. Suppose we consider it, as, loosely speaking, a tool for the collective production and editing of knowledge, by technically untrained people. To some, this knowledge as being represented as triples. Others view it as being represented in the text itself (hence, not really processable), and still others may see it as a combination of both. The first alternative does not really look like a wiki as most people would think of it; I suspect the third one is the more common understanding of what a "semantic wiki" would be. (Btw, I don't see such an advantage to regard a wiki as simply a "text based" interface to directly edit RDF or OWL ontologies... but this is another discussion perhaps). I can't see how to analize advantages/disadvantages of any of the alternatives before it is clear which paradigm is being followed, If you take the first point of view, I'd tend to agree with Mark's remarks. If you take the third point of view, it is not so clear... This is essentially why I asked Mike to make the usage scenarios a bit more explicit; I'd like to understand better how is the formal (i.e. triples) knowledge is being created, edited AND USED in the first and third alternatives above. So, in summary, what is (more precisely) the problem being addressed in using the wiki? On 26/4/2007 19:39, Mark Greaves wrote: > Mike, > > >> Now I have a triple. I don't care where it is stored, it can be >> associate with any page or no page. In fact I don't even want >> to see it. I want the tool to take care of all that for me. >> > > I disagree with this. In order for triples to enjoy the benefits of > social editing and the social identification and correction of errors, > they have to be simple to find, examine, and edit by a large number of > relatively untrained people. In order to maximize the number of people > who can access/edit the triples, the process of locating and editing the > triples needs to be as parallel as possible to the already-known process > for making corrections to ordinary wikitext. So, rather than force > triple-editing to go through some kind of searchbox interface, it makes > more sense to me to make the triples embed in the wikitext of the > subject page. Furthermore, this strategy allows for a natural way of > using associated wikitext to lay out arguments, in case there is dispute > over the value of a triple. > > This does make the kind of freeform triple entry you desire a bit more > cumbersome. Nevertheless, I think it is consistent with the goal of > making the triples that exist as accessible as possible to the wiki > editors. > > Mark > > Mark Greaves > Vulcan Inc. > 505 Fifth Ave S, Suite 900 > Seattle, WA 98104 > > (206) 342-2276 (voice) > (206) 342-3276 (fax) > > > > >> -----Original Message----- >> From: Blanchard, Duane L [mailto:Dua...@bo...] >> Sent: Thursday, April 26, 2007 11:10 AM >> To: Uschold, Michael F; Daniel Schwabe >> Cc: ba...@ba...; Kelly Jones; Jones, David H; >> sem...@li...; Murray, William R; >> sw...@ai...; Mark Greaves; >> csh...@vi... >> Subject: RE: Creating Triples Anywhere in a Semantic Wiki '' >> >> See additional inline comments.... >> >> Thx, >> >> D >> >> ---- >> Duane L. Blanchard >> Computational Linguist >> Phantom Works - Mathematics & Computing Technology >> 425.373.2800 >> >> To pick up on this topic, Mike, could you clarify then what >> you have in >> mind when you think of a "wiki page"? >> >> MU: good question. I guess I mean when I click on a wiki concept, like >> say "Bob" and now I'm on Bob's page. Or I might be on the page for the >> wiki concept: Idaho. >> >> DB: Same here, but perhaps the question should be rephrased as what is >> the relation between a wiki page and an entity or a triple in >> your mind. >> -- >> >> Is it just a place to hold an exchange of information (analogous to a >> "thread" in a forum)? If so, what does it mean to attach an >> attribution >> (a triple) to this page - >> should I interpret that all triples have the page as subject, >> which is a >> resource in some sense, and the meaning is whatever you >> assign to these >> triples? >> >> >> MU: this is what Semantic Media Wiki assumes, as I understand it. That >> is why in the triples in the markup, you don't have the subject >> explicitly there, it is assumed. In which case the page only >> really has >> doubles there explicitly, not triples. >> >> This assumption makes it impossible to add a real triple that >> relates to >> something else. >> >> DB: This is correct for Semantic Media Wiki. Each page is a >> concept and >> each relation on that page uses this concept as it subject. >> >> DB: Mike, more please on attaching a triple to a page, or to the wiki >> but not to a page. >> -- >> >> Is it supposed to encapsulate some kind of concept or set of >> concepts - >> then the triples be interpreted as statements about these concepts? >> Must it be in some sense "self-contained" - relative to some >> discourse, >> or relative to some conceptualization schema? >> I'd be very interested in knowing more the requirements (and >> scenarios) >> you seem to have in mind when you state that "that's why I stopped >> trying to use the tool"... what kind of applicaton do you need? >> >> MU: I want to create a wiki page, classify that page as being an >> instance of some class. >> So say I create a page for Bob, he is an instance of Person. On that >> page, I want to be able to write unstructured text. Like Bob lives in >> Idaho. And to be able to create a triple that says that. And I also >> mention Montana in the text, and while I'm thinking about it, Idaho >> borders on Montana. So from the web page, I want to be able to create >> Idaho and Montana as wiki concepts. Then I want to select >> each of them, >> and then choose among a set of relationships which includes bordersOn >> and set that relationship between these two concepts. >> >> Now I have a triple. I don't care where it is stored, it can be >> associate with any page or no page. In fact I don't even want >> to see it. >> I want the tool to take care of all that for me. >> >> Furthermore, the system now should know that Idaho and >> Montana are say, >> regions from the domain and range constraints of bordersOn**. >> >> ** Lets ignore the fact that a state is a political entity, and has an >> associated region. That's an ontology issue, not a tool issue. >> >> DB: Mike, it seems, from my limited exposure, that this is how Visual >> Knowledge operates. If you create a triple, but don't >> associate it with >> any page, how do you later make changes to that triple? In >> SemMedWiki, I >> would go to the page that is the implicit subject and make the change >> there. >> >> DB: In VK, are only pages instances of classes, or can instances exist >> without also being pages? >> >> > > > |