- From: Kingsley Idehen <kidehen@openlinksw.com>
- Date: Mon, 18 May 2009 10:57:06 -0400
- To: David Huynh <dfhuynh@alum.mit.edu>
- CC: Sherman Monroe <sdmonroe@gmail.com>, Linked Data community <public-lod@w3.org>, semantic-web@w3.org
David Huynh wrote: > Sherman Monroe wrote: >> Kingsley wrote: >> >> There are half a dozen Entities across N graphs in the Quad Store. >> The UI issue here is that we don't show the source Graphs in the >> results page. Reason, we know we can actually provide distinct >> results cheaper than listing the Graph Names etc.. >> >> Your timing is borderline impeccable, we will actually be >> releasing the Distinct optimization that showcases what I mean. >> Anyway, for now, when you select one of the Microsofts from >> DBpedia graphs, click on the "Stats" link, it will give you a back >> door view of where the data has come from. >> >> >> David, also keep in mind that this is one of the benefits of >> set-based browsing. If I have a results set of twenty synonymous URIs >> representing Microsoft, for someone researching Microsoft, such a >> list would be a goldmine, because I can click razorbase -> >> Information and view all information for all 20 versions of that one >> entity simultaneously. In fact, when I get sparse results for an >> entity, I always click the razorbase -> Information -> "Alternative >> Identities" (e.g. owl:sameAs) and also I look under the reverse >> properties for the same, to pull any alias that I may not be away of. > Hi Sherman, > > I guess I'm just trying to close the gap between Google's search > results--which people are familiar with--and razorbase's or any novel > search engine's results. For example, when I search for Microsoft on > Google, the first result not only IS what I want, but also LOOKs like > what I want. I can make the decision to click on it within maybe 1 or > 2 seconds. But what happens when you want an entity associated with pattern: Microsoft that isn't the highly referenced company: Microsoft, in google's document index? The view we have is this: 1. Hook into Google and Yahoo and MSFT for pages and even apply a weighting or our algorithm so that Google|Yahoo|MSFT first page will be the same as ours (* this is deliberately not part of the LOD instance since sponging is disabled for now*) 2. Extend this somewhat popularity skewed algorithm with the ability to disambiguate using Entity Type ( Category re. Sherman's UI) and Properties (Information re. Sherman's UI). > The URL "www.microsoft.com" in that search result is perhaps the most > convincing element, as I know only *the* Microsoft can possibly own > that domain. (This will be a challenge for any SW search engine, > because no-one can own any URI, and so, seeing a URI alone means > pretty much nothing. URI means: here is the Identifier for an Data Object in this Data Space. It means that a simple click will unravel its essence in a representation overtly or covertly negotiated between user agent and data server. > That's one of the main differences between URL and URI, which is > usually swept under the rug.) I hope I've taken it out from under the rug. A URL is a URI :-) > > I believe that these "little details" play a big role in how users > interact with SW search engines. It's not just that the search engine > should return the right result, but that it should convince the user > that the right result is right. > Sure, and hopefully, we will collectively make all of these things clearer :-) Kingsley >> As always, enjoy chatting with you :-) >> >> >> Indeed, the LOD owes a lot to your work!! > > Glad to have helped! :) > > David > > > -- Regards, Kingsley Idehen Weblog: http://www.openlinksw.com/blog/~kidehen President & CEO OpenLink Software Web: http://www.openlinksw.com
Received on Monday, 18 May 2009 14:57:53 UTC