Comments on: Namespacing & Context - ramifications for the semantic web

By: Manuel Simoni

Manuel Simoni — Fri, 18 Apr 2008 14:36:38 +0000

My recent thinking re “semantic” data is that maybe a search engine is a better model for such systems than a triple store, which resonates with the points you are making.

A triple store requires very exact inputs to deliver any data at all. In contrast, a search engine’s motto is to “deliver results no matter what”.

Basically, indexing and (page-)ranking all “title literals” (wikinames, blog titles, …) should do the job, you probably wouldn’t need a fulltext search engine.

And let’s not forget the R in REST - a triple store has to return triples, whereas a search engine can return any representation, even a movie clip, which is a much better model for user facing systems.

By: Danny

Danny — Fri, 18 Apr 2008 18:17:41 +0000

Phil, should be an interesting experiment, looking forward to hearing how it goes (it sounds similar to the old Semantic Network stuff, which worked well for conceptual map kind of things, but not so good for practical data kind of applications).

One question - a Semantic Web without URIs in the RDF, where’s the Web bit? That an agent can follow it’s nose (do a GET) to find more information is a feature, no?

Manuel, the triplestore/search engine is a bit apples/oranges - most search engines have some kind of database under the hood, the database itself will require fairly exact input.

Going the other way, a triplestore with SPARQL can make a very simple backend to a search engine, e.g. to find anything with the word “primer” in it:

SELECT DISTINCT ?o
WHERE {
?s ?p ?o .
FILTER regex( ?o, ‘primer’, ‘i’ )
}
LIMIT 10

try pasting it in here:
http://hyperdata.org/sparql/demo/sparql-editor.html

By: Eric Jain

Eric Jain — Sat, 19 Apr 2008 07:26:05 +0000

Don’t know about doing away with URIs… But some kind of context seems essential when looking at data from different sources. Two resources with different URIs may be the same thing, and two resources with the same URI may be different. Moreover, the meaning of “same” itself is context dependent!

By: Richard Cyganiak

Richard Cyganiak — Wed, 23 Apr 2008 11:24:28 +0000

As I see it, URIs in RDF are 10% about disambiguating between terms, and 90% about allowing lookups of terms.

I suspect that by getting rid of the 10%, you also lose the 90%.