Re: totalitarianism

* Jeremy Carroll <jjc@hpl.hp.com> [2004-05-28 10:37+0200]
> 
> Reading the earlier thread 'Comment on "Meaning and the Semantic Web"', 
> my mind came to the following (rather old) example.
> 
> 
> "This song is dedicated to the World Health Organization, it's a medical 
> song and it concerns a disease, whose classification according to the 
> Internationational Classification of Diseases is 
> three-oh-two-point-zero."
> Tom Robinson, intro to "Sing if You're Glad to be Gay" 1978
> 
> Bijan and Peter:
> > Even if so, we believe we have
> > embraced only those portions of anarchy that are necessary to prevent
> > totalitarianism, for any proposal for Semantic Web meaning
> > that cuts off easy access to disagreements will inevitably end up
> > stultifying the Semantic Web.
> 
> John:
> > I am finding this reference to totalitarianism hard to accept.
> 
> Apparantly factual scientific classifications, such as the one that Tom 
> Robinson referred to, can convey ways of thinking and assumptions that 
> are oppressive and unwelcome (at least to some).
> Tom Robinson, using natural language, could refer to this scientific 
> classification and then launch into a bitter critique of it from his 
> experience. The ability of natural language to refer to other people's 
> concepts without agreeing with them, or with only partial agreement is 
> the basis of dissent, which is the motor of social change - which at 
> least in some cases is for the better (e.g. IMO, in this case).
> 
> Because of the monotonic discipline of the Semantic Web it is very hard 
> to unsay something: if in referring to something like "Disease" we have 
> to effectively assent to someone else's definition, which could will 
> include a complete classification, then the formal theories do not 
> allow us to unsay part of that definition.
> 
> While the imports mechanism that Bijan and Peter suggest is perhaps not 
> enough to fully address this problem, the concept of URI ownership 
> could be as dangerous as allowing the WHO to own the word "disease", if 
> word's could have owners.
> 
> I think "totalitarianism" was an appropriate choice of word, as argued 
> particularly by Peter. A traditional means of totalitarian regimes to 
> first gain hold is to control the press (before television). 
> Publication technologies that allow central points of control should be 
> avoided.
> 
> Of course more up to date examples would be more controversial e.g. 
> classifications of marital status, or of wars.
> 
> It seems well within scope of the SW that deployed concepts should be 
> open to critique, revision and change.

Nice example. Here's another, drawn from Wordnet. Wordnet is a lexical
database whose noun hierarchy (aka hypernyms, modelled on "a ___ is a kind of 
___" template) can just about be used as basis for an RDF/OWL class 
hierarchy.

There are some hierarchical relations in Wordnet which may well reflect 
attitudes that are evident in English usage, but which --- if we turn it 
into an Ontology --- would get "taken literally" by RDF tools.

danbri@snowball2:~$ wn  nymphomaniac -hypen

Synonyms/Hypernyms (Ordered by Estimated Frequency) of noun nymphomaniac

1 sense of nymphomaniac                                                 

Sense 1
nymphomaniac, nympho
       => pervert, deviant, deviate, degenerate
           => reprobate, miscreant
               => wrongdoer, offender
                   => bad person
                       => person, individual, someone, somebody, mortal,human, soul
                           => organism, being
                               => living thing, animate thing
                                   => object, physical object
                                       => entity
                           => causal agent, cause, causal agency
                               => entity



Apologies for the text formatting, this is just what the Wordnet 
commandline tool emits. 

You can also some some other interesting bits of the hierarchy by
searching on terms like 'nigger', 'queer' etc. The 'nympho' example 
is perhaps more interesting because the more offensive terms are higher 
up the tree, and so could be inferred by class reasoning from the 
leaf term.

Incidentally, this illustrates some issues with moving too rapidly from
a lexical database into an ontology. The former is 'words people use',
the latter 'the way the world is'. We would do well to always emphasise 
that ontologies capture a worldview and that those worldviews may be 
highly contested.

Dan

Received on Friday, 28 May 2004 04:54:41 UTC