RE: sameas.org

>Hugh Glaser wrote
>>On 05/06/2009 04:09, "Giovanni Tummarello" <g.tummarello@gmail.com> wrote:

>>>> a New Zealander and a Kiwifruit)
>>>> 
>>>> throws up a radio station, an animated cartoon and lots of wordnet links to
>>>> a
>>>> juggle of plumbing but no juice.  No sign of
>>>> 
>>>> http://dbpedia.org/resource/Kiwi however
>>> Ah.
>>> We only look at the first n results from Sindice, and clearly kiwi is a
>>> popular name.
>>> Clicking on the sindice link will show you what we got.
>>> However, I see that using "kiwi bird" gives a decent return, so all is not
>>> lost.
>>> Best
>>> Hugh
>> 
>> 
>> we definitely have to improve here, luckly we know what's wrong and
>> what's next and its coming. the ranking works decently for some
>> entities but less for others.
>> 
>> a major fix of this hopefully before summer break
>> Giovanni
>> 
>Not sure it needs much "fixing", but I guess things can always be improved.
>I can immediately think of 3 or 4 non-bird Kiwi meanings that I would expect
>to come above the bird.
>It's a great example of why SW/LD helps us, and we can do things like
>Kingsley's post.

I think what I was getting at is that searches in the semantic web might be expected to order results semantically rather than in random order, or in order of current popularity as in non-semantic search engines. I expect the bird to come first without the need for user intervention to refine the search with additional words or type filters because this is the initial meaning of the word in Maori, and all other usages are derived from it, some later than others.  

Clearly as the size of the LOD cloud increases, the problem of ordering results becomes more important, especially as services place limits on the number of results returned.  I realize that deriving a measure of semantic distance is a really hard problem but in a way, isnt it rather central to the semantic web?

Kingsley, I tried http://lod.openlinksw.com, but here too, when searching for, say Bristol, I get pages of people who mention Bristol in their foaf profile, and things which are parts of Bristol (Bristol Airport) and hundreds of other mentions whilst the primary meaning of the word is quite lost. That just doesn’t seem like the expected behaviour of a 'semantic search engine'.   I wonder for example if something could be done with dates to get the results into a more semantic order?

Chris Wallace


This email was independently scanned for viruses by McAfee anti-virus software and none were found

Received on Friday, 5 June 2009 10:46:12 UTC