Queries Re: RDF IG meeting at W3C Technical Plenary 2001-02-28 (fwd)

Having subjects that match the content is kind of important to allowing
people to track data.

Chaals

-- 
Charles McCathieNevile    http://www.w3.org/People/Charles  phone: +61 409 134 136
W3C Web Accessibility Initiative     http://www.w3.org/WAI    fax: +1 617 258 5999
Location: I-cubed, 110 Victoria Street, Carlton VIC 3053, Australia
(or W3C INRIA, Route des Lucioles, BP 93, 06902 Sophia Antipolis Cedex, France)

---------- Forwarded message ----------
Date: Mon, 12 Feb 2001 09:19:48 +0000
From: David Allsopp <dallsopp@signal.dera.gov.uk>
To: www-rdf-interest@w3.org
Subject: Re: RDF IG meeting at W3C Technical Plenary 2001-02-28
Resent-Date: Mon, 12 Feb 2001 04:21:53 -0500 (EST)
Resent-From: www-rdf-interest@w3.org



Seth Russell wrote:

> > If we express queries in RDF, and parse them into a Model, how does the
> > query engine access them in order to carry out the query?  By querying
> > the model...?
> >
> > [ OK, I guess you'd probably only need Model.find(s, p, o), but the
> > circularity amuses me. 8-) ]
>
> Obvioulsy any angent who might answer the question needs to know it is a
> question.  In the example below I think I have adequeately identified that
> situation.

I meant, how does the agent extract the individual property values from
the graph which expresses the query, since that would require carrying
out a query, but it was a flippant point anyway 8-).

> > My first concern is that some of the proposed queries expressed in RDF
> > are pretty much unreadable, and extremely verbose, at least in M&S
> > syntax, which may partly explain why people tend to go for SQL-like
> > query syntax, which, after all, just expresses a graph with variables in
> > a cleaner way.
>
> Actually I think it falls on the other side.  Putting the variables (via
> pronouns) directly in the RDF will be far more readable and far shorter - your
> welcome to take a stab at my example below in your favorite query language.

This is assuming that the community will adopt a syntax like N3 or
SemEnglish, presumably? I which case I am satisfied; I'm just appalled
by the thought of having to write (or read) complex queries in M&S.

> Bear in mind that to write a query you must know the exact property names in the
> destination model that might yield the answer.  This will not normally be the
> case for practical applications if you factor in the diversity of the semantic
> web.  The person asking the qusetion knows only the definition of the question
> in their terms;  they lack the  knowledge of how the answer might be expressed
> inside of your internal system;  therefore they will not normally have enough
> information to write a query to you data base.

If we don't have shared terminology, how can either agent possibly
understand the query or the reply? Isn't that the whole point of
ontologies, namespaces, etc?

> We need to define a process
> between the question and the query where translations of property equivaentTo(s)
> and inverseOf(s) and other kinds of inferences happen; and/or the question is
> servered up to a human agent in a language that is comprehensible.

Ah, I see - we may not initially share all the necessary terminology,
but we must achieve this before the query can be satisfied. Fine.

At the moment, I'm chiefly concerned with more tightly defined agent
systems (not as much diversity as the web as a whole) where the
ontologies are mostly known in advance, and new ones are integrated
fairly explicitly (by adding translator agents who know the various
equivalentTo properties, for example).  I'm not sure what your
'natural-language-style' queries add in this situation, where very exact
queries are required (obtaining a piece of technical data for a
aircraft, for instance) - whatever process you use, one precise query
has to be mapped to another precise query, or it cannot be satisfied.

> > (I've been very struck by the similarity between the SQL-like queries,
> > and N3 - I had a spooky feeling when I was playing around with some RDF
> > query syntax like N3, then on the same day found the pages on N3 for the
> > first time!)
>
> If you liked N3, you'll love SemEnglish (see below).  Note that is can be
> legitimately embedded in XML, is cleaner and easier to read and write.
> Incidentally, my signature example asks a real question ... the answer to which
> I really do want to know.  (hmmm.... i wonder if i can say that any plainer).

Looks pretty clear to me! If I was being _really_ pedantic, I would say
that it isn't a question; it's a statement that Seth is asking a
question. 8-)

> <signature
>        language="SemEnglish"
>        fromDictionary="http://webns.net/roughterms/#"
>        context="ThisEmailTrain"
>        intent="toRecieveAnswer">
> Seth
>      preferedEmail "seth@robustai.net";
>      properName "Seth Russell";
>      asks [Who#x willProgram Sembrowser].
> Who#x
>      worksWellWith Seth;
>      fluentIn RDF, N3, VB, VFoxPro, "C++", SQL;
>      canEasilyLearn BerkeleyDB;
>      salaryRequirements "0".
> BerkeleyDB
>      see [http://www.sleepycat.com/].
> Sembrowser
>      description "Semantic Web Browser";
>      aka MyMemory, ContexBrowser, Emouth, CyberMind;
>      willBeLike [Moasic wasTo Www].
> SemEnglish
>      see [http://robustai.net/mentography/semenglish.html].
> </signature>

Regards,

David Allsopp.

-- 
The peason-molesworth space ship threatened by wild mercurian maths
masters lies disabled in the onion bed of the kitchen garden.
Meanwhile in the master's common room, Sigismund arbuthnot the mad
maths master musters his rhomboids...

Received on Monday, 12 February 2001 06:56:33 UTC