Join GitHub today
GitHub is home to over 28 million developers working together to host and review code, manage projects, and build software together.Sign up
Qualifying validity of semantic properties #985
I'd like to open a discussion on a potential new feature for SMW that would require some work. I recently posed this question on the SMW-user mailing list, but figured it might be better to move the conversation here so any progress might be tracked. I have modified this post from my original email based on a couple email replies.
By implementing SMW, we moved our "static" knowledge base from Word documents, emails, and other traditional file types into a much more powerful semantic wiki. Now we can make multiple use of the same information and we can query for data to expose gaps in our knowledge capture. This is all great as we know. The limitation, though, is that our organization needs to back up many of these semantic "claims" with some kind of source.
For example, I can say for the "EVA Ratchet Wrench" page (Subject) that the property "Length" (Predicate) has the value "12 inches" (Object). The problem I see is that I have to use another method or extension to link to that statement the source data for that claim. In this case, I should link to the drawing for that hardware to verify that it is indeed 12 inches long. Otherwise, the only validation we have is the record of who added the value of "12". How do we know they didn't look up the wrong number or have a typo mistake? In the wiki way, we rely on other users reviewing edits for accountability. But what source do they use to verify the edit was correct?
We use Extension:Cite to help us back up our claims with sources like hardware certification documents and operational constraints (one of the many types of rules we have to follow). Unfortunately, this extension doesn't work well with SMW (or maybe I should say it the other way around).
So why not add a fourth element to the relationship? Could (and should) SMW be modified as Subject - Predicate - Object - Source?
I suppose it might not be that simple, though. My example of the length of an object is pretty easy to source. Something like the population of a city might not be so easy. For example, the population of Paris might be set to 2,273,305 with a source based on data from 2013, but that is obviously outdated. We might also make a claim that the population is 2,400,000 based on some extrapolation. Which of those two statements is more "true"?
Maybe Wikibase is the solution to this problem, but I'm still trying to figure out how well it will integrate with SMW.
So back to my original intent -- I'm wondering if SMW could be modified to allow for linking the data source(s) to make citations a little better integrated.
The following comment will focus on the reference made to the
Given the fact that we want structured data in the first place we need a different approach. The following sneak preview  of the new
Thanks for the preview. I'm excited to try it out. In the video, it looks like you can establish the reference information on a separate page from where it is cited. Is that true? If so, can it be set on any page or just the subpage named
Also, since the code to cite a reference is
I'm not sure about the syntax I used, but basically I'm asking if you can set something as a semantic property and link it to the cited source data. In my example, with the Population definition next to the CiteRef definition, I don't see how they will be linked. If I query from a different page for the population, will it also bring up the citation? For example, would
Yes (that was one main motivation, reusable citations)
Yes (those enabled for semantic usage)
As I tried to make it clear in the pretext of the previous comment, we talk about plain simple citation (no qualification of statements) therefore those will not be linked to each other. Qualification of statements is a conceptional model issue in SMW and not solved by adding a simple link.
In above case, you get a claim about the
Just as a brainteaser you would need a different syntax that would allow to combine two statements using an intermediary object and make them link to each other (but that goes beyond the current citation discussion).
No. Again this needs more food for thought and is not solved by adding citations to a subject.
PS: The main motivation for
referenced this issue
Aug 24, 2016
SMW-core did require some internal changes to allow recording of meta data (or as we called provenance data) to an individual value assignment.
#1808 tries to outline a nominal approach that should be sufficient for users to adapt individual requirements to store and query those data.
I'll be closing this issue in favour of #1808.