[MOBY-l] Fwd: Problems searching with semantic moby dot org

Tue Aug 24 23:25:27 UTC 2004

Frank,

Sorry, but the semanticmoby.org web site is still pretty rough, hence 
the "Under Construction" banner needs to be taken quite literally for 
the time being. I have about three weeks worth of work on Semantic MOBY, 
running locally, that has not yet been released to semanticmoby.org; 
things should be in a much better state by the end of the week.

That said, here is some (long-winded) clarification - I need to add 
similar explanation to the site itself. Semantic MOBY does indeed use an 
RDF-based discovery and engagement paradigm (post the RDF/XML 
serialization of a query graph to the discovery server and get back a 
set of matching graphs as RDF/XML; post an RDF/XML ("input") graph to a 
Semantic MOBY provider and get the RDF/XML graph back (as "output"), 
where the graph has been augmented by having blank nodes filled in). 
This really is the heart of Semantic MOBY, and it needs to be (will be 
as soon as I can get to it) documented and made accessible.

However, RDF/XML is very unwieldy for non-geek users. Hence, one of the 
goals of Semantic MOBY was to have a "Google-like" interface for 
discovery and engagement Semantic MOBY service providers. That is what 
you see at the www.semanticmoby.org home page. In a sense, it's the 
Semantic equivalent of the MOBY Central 'gbrowse' front end 
(http://mobycentral.cbr.nrc.ca/cgi-bin/gbrowse_moby) and the very 
immature beginning of a "Semantic Google".

In a nutshell, here is how this works: when the Semantic MOBY discovery 
server decides to add a provider graph to its repository of known 
services, it looks at all the nodes in the graph that are of type 
moby:Provider, moby:Subject, and moby:Object. For each of these nodes, 
it looks for other statements asserting the node to be of other types 
(RDF nodes can have multiple rdf:type statements made about them). For 
each of these additional types (i.e. the objects of the rdf:type 
statements), the discovery server retrieves the class definition (to be 
a good MOBY player requires that all classes are retrieveable by HTTP 
GET - see, for example, the Semantic MOBY core definitions at 
www.semanticmoby.org/ontologies/core), and looks for moby:keyword 
property statements in these class definition. For each such keyword, 
the discovery server remembers that the provider node (or subject node 
or object node) in the provider graph is associated with that keyword.

When you do a search from the www.semanticmoby.org page, the discovery 
server looks in its stored keyword/provider associations in order to 
find providers that match "by keyword" and displays links to them in 
Google-like fashion. That gets us to the provider engagement side of the 
equation, which would take at least as long to describe as the 
description that I just gave about discovery, so I'm going to skip it 
for now, other than to say that the links are to another Semantic MOBY 
service which tries to invoke and display the results of invoking the 
provider.

This "keyword based" search is completely separate from the Semantic, 
RDF-graph based search, which is also there (but again, not yet 
documented). We (Damian and I) decided that the most conspicuous, hence 
first thing you see on the site, version of Semantic MOBY search should 
be up front; the RDF-graph based search, though in a sense more what 
Semantic MOBY is all about, will be less conspicuous, to be dug up by 
the "real geeks" such as readers of this list :-)

Regarding the "Not Found" error you got, I believe that the Servlet 
container had failed to restart when the server got rebooted the other 
day - sorry. It is up now, but again, a lot of the functionality that I 
described is still under construction, and should be there by the end of 
the week...

// Gary

Frank Gibbons wrote:
> Mark,
> 
> For some reason, the entire body of my recent posting was stripped. I 
> also had a message telling me that the message would be subject to 
> moderator approval (perhaps because of the included HTML?). I'm 
> re-sending in plain text.
> 
> -Frank
> 
>> Date: Tue, 24 Aug 2004 16:35:05 -0400
>> To: moby-l at biomoby.org
>> From: Frank Gibbons <fgibbons at hms.harvard.edu>
>> Subject: Problems searching with semantic moby dot org
>>
>> Hi,
>>
>> I've tried using semanticmoby.org to see what's out there in the 
>> S-MOBY world. I know it just went up a few weeks ago, so it's bound to 
>> have some hiccups. Several things struck me immediately, and I thought 
>> you'd like to hear about them:
>>
>> 1. How do I compose a query? There is a link to 'Help on Searching', 
>> but it seems curiously generic ('HELLO means the same thing as hello 
>> as does HeLlo"), not at all specific to the semantic task at hand. 
>> Also, it appears that the page works by allowing the user to specify a 
>> provider, subject and object of interest, but isn't the real thing 
>> supposed to query using an RDF graph of subject-predicate-object? How 
>> far off is that?
>>
>> 2. A couple of examples would help a lot, both for the current 
>> (provider-subject-object) incarnation, and especially when the full 
>> RDF capabilities are added. I believe this is true even if the 
>> services that are 'discovered' don't do much that's practical yet.
>>
>> 3. When I click the 'search' button, this message appears:
>>
>>
>>
>> Not Found
>>
>>
>>
>> The requested URL /keyword-search was not found on this server.
>>
>> ----------
>> Apache/2.0.49 (Unix) DAV/2 Resin/3.0.8 Server at semanticmoby.org Port 80
>>
>> Just my 2 cents.
>>
>> -Frank
>>
>> PhD, Computational Biologist,
>> Harvard Medical School BCMP/SGM-322, 250 Longwood Ave, Boston MA 
>> 02115, USA.
>> Tel: 617-432-3555       Fax: 617-432-3557       
>> http://llama.med.harvard.edu/~fgibbons
> 
> 
> PhD, Computational Biologist,
> Harvard Medical School BCMP/SGM-322, 250 Longwood Ave, Boston MA 02115, 
> USA.
> Tel: 617-432-3555       Fax: 617-432-3557       
> http://llama.med.harvard.edu/~fgibbons
> _______________________________________________
> moby-l mailing list
> moby-l at biomoby.org
> http://biomoby.org/mailman/listinfo/moby-l
> 
> 