W3C home > Mailing lists > Public > semantic-web@w3.org > May 2009

Re: Yet Another LOD cloud browser

From: Sherman Monroe <sdmonroe@gmail.com>
Date: Fri, 15 May 2009 16:45:15 -0500
Message-ID: <e23f467e0905151445w5637621dv592b1ba1502439d7@mail.gmail.com>
To: David Huynh <dfhuynh@alum.mit.edu>
Cc: Linked Data community <public-lod@w3.org>, semantic-web@w3.org
David,


>
> So, I typed in "Microsoft" and got to
>   http://lod.openlinksw.com/fct/facet.vsp?cmd=text&sid=60306
> Which doesn't look like a permanent link for referring to the query I just
> typed. Copying and pasting that URL to a different browser yields
>
>   An unexpected error was encountered while processing your request.
>   Diagnostics
>
>   SQLSTATE: 22023
>
>   SQLMSG  : SR016: Function length needs a string or array as its argument,
>   not an argument of type 189 (= INTEGER)
>
> But anyhow, it's not a big deal.



In addition, I've added a "link to this" link on the green bar on
razorbase<http://www.razorbase.com>.
Warning, I purge the saved queries on system reboots.


>
> My next question is, how does a person know which one among those results
> is *the* Microsoft? There are half a dozen dbpedia:Microsoft, so which one
> is the one? And if they are all the same thing, why are there multiple
> copies? Is it a glitch in the browsing engine or a glitch in the data?


I think good old fashioned ranking could help with this problem. E.g. a
naive but plausible approach would be to rank URIs based on the number of
triples in which they participate (particularly as the Object). The Virtuoso
Entity Rank is a good example of this.  If you type "Microsoft" into
razorbase (be sure the 'named' option is selected), then you get tons of
things with Microsoft in it's name, but Microsoft Corp. comes up as the
first result. Ditto for other types of named entites.



> Also, I suspect that a random user might not be aware of dbpedia at this
> time, so perhaps "dbpedia:Microsoft" might not sound like *the* Microsoft,
> at least without any further detail. Maybe one copy of  "Microsoft
> Corporation (source: Wikipedia, data governance: Dbpedia)" might sound more
> comprehensible?



That's a data problem :) The hope is that, as we build UIs that expose the
value of linked data, people will be better about being sure their instance
data has the types of annotations (e.g. rdfs:label) that helps us help users
find what they need. In addition, I believe better RDFizing methods will
help.



> Better yet, an image of Microsoft logo next to that would make things
> crystal clear. Even better would be to see the description, "Microsoft
> Corporation is an American multinational computer technology corporation,
> which rose to dominate the home computer operating system market with MS-DOS
> in the mid-1980s, followed by the Windows line of operating systems." below
> it.
>


Hmm, I tend to think that a sparse UI is better for the user whose goal is
to jump from nothing to a complex set of things as quickly as possible, as
user transverses the 'sets', the more text onscreen for user to scan, the
more drag in introduced. I have no data to back up that assumption, and this
is just one mode of browsing among several. I do know that in usability
tests with friends/family for the earlier versions, 0/all people failed to
glance at onscreen notes and descriptions of controls, and instead abandoned
their task instead of reading help text. So I try to keep the data presented
to a minimum, I display URI under title or results in hopes that it gives a
hint of what the thing is and it's source.


-- 

Thanks,
-sherman

I pray that you may prosper in all things and be healthy, even as your soul
prospers
(3 John 1:2)
Received on Friday, 15 May 2009 21:45:54 GMT

This archive was generated by hypermail 2.3.1 : Tuesday, 26 March 2013 21:45:29 GMT