- From: Harry Halpin <hhalpin@w3.org>
- Date: Tue, 18 Jan 2011 16:58:07 -0000 (GMT)
- To: "Michael Hausenblas" <michael.hausenblas@deri.org>
- Cc: "Harry Halpin" <hhalpin@w3.org>, "RDB2RDF WG" <public-rdb2rdf-wg@w3.org>
> > Harry, > > >> While this does not address all Linked Identifiers, my proposal allows >> R2RML >> to at least re-use the results from search engines. > > This is a simple and effective proposal and I appreciate it. But one major > issue remains, and before we go down this route I'd like to see some > clarification on how you plan to address this. > > Note: my experience is of practical nature with systems like Dbpedia > lookup > or the Uberblic API and not so much of theoretical background as obviously > Harry has. > > As long as the thing you're looking up is unambiguous, no problem. > However, > what if the query (and here comes the old example again, sorry ;) looks as > follows: > > http://www.example.com/r2rml?string=Paris > > Now, do you mean the capitol of France? Or maybe a more or less well-known > D-class celeb and offspring of the Hilton clan? > > http://www.example.com/r2rml?string=Paris&results=3 > > ... would, btw, make no big difference, the burden is still on the > end-user > to figure out which is the 'right' entity. Well, that's why there should be a sort of "I feel Lucky" button where you just hope for the best (i.e. a setting in R2RML), but otherwise you would want some kinda of GUI to display the URIs for you. We can't specify the GUIs, but we can specify that there should be someway for the user to go to manual select mode. Usually ambiguity can also be sorted by some kind of query expansion, i.e. you can query not just the string in the key, but all the strings in the database that are associated with that row/column. See my Semsearch 2010 paper with Lavrenko for results there - we went up from 60%->90% accuracy on real world queriees. My experience with Semantic Web search engines at the Semsearch evaluation last year, which I helped co-ordinate, was that they are actually quite terrible. But they should get better :) > > Thoughts? > > Cheers, > Michael > > -- > Dr. Michael Hausenblas, Research Fellow > LiDRC - Linked Data Research Centre > DERI - Digital Enterprise Research Institute > NUIG - National University of Ireland, Galway > Ireland, Europe > Tel. +353 91 495730 > http://linkeddata.deri.ie/ > http://sw-app.org/about.html > > > >> From: Harry Halpin <hhalpin@w3.org> >> Date: Tue, 18 Jan 2011 16:28:54 -0000 (GMT) >> To: RDB2RDF WG <public-rdb2rdf-wg@w3.org> >> Subject: Re: ISSUE-7 (linked-data): Linked Data aspects for mapping >> RDB2RDF >> [R2RML] >> Resent-From: RDB2RDF WG <public-rdb2rdf-wg@w3.org> >> Resent-Date: Tue, 18 Jan 2011 16:29:57 +0000 >> >>> >>> ISSUE-7 (linked-data): Linked Data aspects for mapping RDB2RDF [R2RML] >>> >>> http://www.w3.org/2001/sw/rdb2rdf/track/issues/7 >> >> While this does not address all Linked Identifiers, my proposal allows >> R2RML >> to at least re-use the results from search engines. >> >> http://lists.w3.org/Archives/Public/public-rdb2rdf-wg/2010Dec/0026.html >> >> I have to miss the call as well, but if this goes on the agenda I'd be >> happy to talk about it in detail next call, particularly if it gets some >> review, and would then feel confident in fleshing it out. >> >> cheers, >> harry >> >> >>> >>> Raised by: Michael Hausenblas >>> On product: R2RML >>> >>> As per our charter [1] "the mapping language MUST allow for a mechanism >>> to >>> create identifiers for database entities. The generation of identifiers >>> should be designed to support the implementation of the linked data >>> principles. Where possible, the language will encourage the reuse of >>> public identifiers for long-lived entities such as persons, >>> corporations >>> and geo-locations." >>> >>> Further, we have the Wiki page [2] that summarizes the requirements and >>> possible attempts to address this issue as well as the requirement as >>> per >>> our UCR doc [3]. >>> >>> [1] http://www.w3.org/2009/08/rdb2rdf-charter >>> [2] http://www.w3.org/2001/sw/rdb2rdf/wiki/LinkedDataAspects >>> [3] http://www.w3.org/2001/sw/rdb2rdf/use-cases/#GUIDGEN >>> >>> >>> >>> >> >> > > >
Received on Tuesday, 18 January 2011 16:58:13 UTC