page rank

[VladimirAlexiev]

(Split from #19)

@chile12 said: I'd suggest a new property like: dbo:pageRankValue = http://purl.org/voc/vrank#hasRank ° http://purl.org/voc/vrank#rankValue (http://www.aifb.kit.edu/images/4/4d/Dbrank2012.pdf)
This will enable us to create page-rank datasets in NT format with just one triple per resource.
Atm it looks like this:

http://dbpedia.org/resource/!!!_(Chk_Chk_Chk) 
  http://purl.org/voc/vrank#hasRank 
    [http://purl.org/voc/vrank#rankValue 
      "0.15"^^http://www.w3.org/2001/XMLSchema#float] .

Vladimir: I'd say simplify the name: dbo:pageRank (range xsd:double).

And how about computing RDF Rank? http://graphdb.ontotext.com/documentation/standard/rdf-rank.html. We've found such measure quite useful for displaying "top" entities satisfying some condition.
Eg try this factforge query

[chile12]

That is the idea. Until now we were provided with the page-rank dataset. Starting with this release we will compute it on our own.

[VladimirAlexiev]

@chile12 To clarify: will your rank be based on RDF resource links, and not only page links?

[joernhees]

i'm not entirely sure what's planned either... both would be interesting, but computing page-rank on (mostly) human generated links (e.g., dbo:wikiPageWikiLinks) is probably more meaningful

[chile12]

We will probably go with: https://github.com/athalhammer/danker by Andreas Thalhammer. Which would need minimal involvement from our side. Please have a look and share your thoughts.

[joernhees]

yepp, i think that works on pagelinks only... https://github.com/athalhammer/danker/blob/master/lib/createLinks.sh downloads the files, resolves redirects etc... probably you have all necessary information already during release and could just re-use them... all sorts can probably be sped up a lot with -S 2G if you can spare 2 GB of RAM

[athalhammer]

Yes, experience shows that the ranks of ontological/organizational concepts (such as male, English, etc.) are skyrocketing in pure RDF settings. This is often not useful for applications such as autocomplete and the like. I'm currently experimenting with mixed settings (see experimental branch of danker) which provide more coverage.

@chile12 I think we can extend the vrank vocabulary accordingly. Let me check whether it is possible to use the old namespace for that.

@joernhees pull requests welcome ;)

Andreas

[athalhammer]

I made some suggestions on how the new property could look like, I would be happy about any feedback on 6020peaks/vRank#1

[athalhammer]

I opened a pull request on vrank for a new release planned for 31.10.2017:

6020peaks/vRank#2

I'm not entirely sure about the property chain axiom and the implications, so reviews are welcome.

[athalhammer]

Dear all,

We just released a new version of vRank (including vrank:pagerank):

https://github.com/6020peaks/vRank/releases/tag/v2.0

Happy Halloween!

[reeshabhranjan]

Hey, is this issue still open? Are there any improvements needed in the existing implementation?