Re: objects to facts to links to LOD to ???

shouldn't trust be part of layer where we link distinct information peices
into GGG ? say in we got information from one source that capital of state A
in country B is C. rather than clubbing this information directly into GGG,
shouldn't we just reconfirm it with other sources and then add it if
majority of sources agree. to build trust at the layer 5 we need some meta
data in additon to the GGG to trust some information in lieu of other.


as far as discovery is concerned, i think it should be mostly in layers
above 3(and some part in layer 3 eg. sameas). can someone share few varied
examples of few discovery scenarios ?





On Tue, Dec 16, 2008 at 10:57 PM, Kingsley Idehen <kidehen@openlinksw.com>wrote:

> On 12/16/08 10:44 AM, Knud Hinnerk Möller wrote:
>
>>
>> Hi,
>>
>> I'm not sure about the discovery part - that seems to be more a meta-layer
>> with data describing linked datasets (VoiD [1]). However, I agree that
>> inference would have to be in layer 4 and up. Things like non-explicit
>> object consolidation, vocabulary and ontology mapping, etc. will be
>> necessary to integrate the data even beyond the point of explicit links.
>>
>> @ravinder: I like the three layers you propose - is there a reference for
>> that, or did you come up with it just now? I'd love to reference it in my
>> PhD thesis!
>>
>> Cheers,
>> Knud
>>
> Knud et al,
>
> I think Ravinder has started the process of fixing the current Semantic Web
> layer cake :-) Which is a very good thing (imho, but not seeking a Layer
> Cake discussion explosion).
>
> The tricky part is the interchangeable nature of "Discovery" and "Trust" in
> any such scheme layer-wise. For instance, do "Discovery" and "Trust" occupy
> Layers 4, 5 or either ? We ultimately want to reason against trusted data
> sources, but the serendipity quotient of discovery is a key factor re. the
> dynamic nature of "trusted sources".
>
> Since I am clearly thinking and writing (aloud) at the same time, I would
> suggest:
>
> Layer 4 - Discovery (with high Serendipity Quotient)
> Layer 5 - Trust (albeit inherently volatile)
>
> Kingsley
>
>
>>
>> [1] http://semanticweb.org/wiki/VoiD
>> On 15.12.2008, at 20:54, Juan Sequeda wrote:
>>
>>  Hi Ravinder,
>>>
>>> Interesting points. I would say that Layer 4 would be "discovery".
>>> Furthermore, this would be an inference layer that would allow discovery.
>>> After having linked data, applications will be able to discover new data. So
>>> maybe layer 4 and up are part of the applications that enable discovery.
>>> Anyways, that is what I see what the semantic web is about: discovery and
>>> serendipity.
>>>
>>> Juan Sequeda, Ph.D Student
>>> Dept. of Computer Sciences
>>> The University of Texas at Austin
>>> www.juansequeda.com
>>> www.semanticwebaustin.org
>>>
>>>
>>> On Sat, Dec 13, 2008 at 11:16 AM, रविंदर ठाकुर (ravinder thakur) <
>>> ravinderthakur@gmail.com> wrote:
>>> Hello Friends,
>>>
>>>
>>> We have now reasonably good repository of linked data but i am not able
>>> to find any good application being made out of it. From the dicsussion on
>>> these mailing lists and other places, i got the feeling that everyone in
>>> semantic world thinks that semantic web is something big but unfortunately
>>> nobody is able to think of an application for _general public_ that can be
>>> based on semantic web data we have currently.
>>>
>>>
>>> This led me to wonder that are there any more layers of information
>>> organization missing in the semantic web stack that are needed to generate
>>> good usable semantic web data which can be more useful for generating useful
>>> semantic apps . Few of the concepts i was thinking for upper layers to
>>> implement were like categorization, set formation(eg MIT passout in 2002),
>>> set intersection, etc. With these higher level layers I was hoping to build
>>> a system to find higher level relations say between places with low cancer
>>> rates and the food people eat there or any perticular gene found in such
>>> people.
>>>
>>>
>>> The current semantic stack looks like this:
>>> .....
>>> layer5 (???)
>>> layer4 ( ???)
>>> layer3 linked facts (isa(mars,planet) AND partof(mars,solarsystem) AND
>>> mass(mars,6566.1255kg))
>>> layer2 facts (isa(mars,planet), isa(cow,animal))
>>> layer1 objects (eg cow, mars, man)
>>>
>>>
>>>
>>> Are there any other thoughts on the need of layers above the layer 3
>>> (linked data ) or these layers will be defined by the respective apps
>>> developers ? Even if there isn't any need, i would atleast like to have a
>>> discussion on the kind upper level layers we might need :)
>>>
>>>
>>> thanks
>>> ravinder thakur
>>>
>>>
>>>
>>>
>> -------------------------------------------------
>> Knud Möller, MA
>> +353 - 91 - 495086
>> Smile Group: http://smile.deri.ie
>> Digital Enterprise Research Institute
>> National University of Ireland, Galway
>> Institiúid Taighde na Fiontraíochta Digití
>> Ollscoil na hÉireann, Gaillimh
>>
>>
>>
>>
>
> --
>
>
> Regards,
>
> Kingsley Idehen       Weblog: http://www.openlinksw.com/blog/~kidehen<http://www.openlinksw.com/blog/%7Ekidehen>
> President&  CEO
> OpenLink Software     Web: http://www.openlinksw.com
>
>
>
>
>