Re: TAP definition of Terrorist

* Rob McCool <robm@robm.com> [2004-06-11 15:04-0700]
> > http://tap.stanford.edu/cgi-bin/download.pl?format=rdfs&iotoo=yes&node=TerroristOrganization&full=yes&download=download
> > http://tap.stanford.edu/cgi-bin/kb.pl?node=TerroristOrganization&op=show&syn=10&browse=all
> > 
> > Resource  > Agent > Organization >
> >    PoliticalOrganization > TerroristOrganization
> > 
> > ...lists 7 organisations as "Terrorist" without giving any definition
> > for this notoriously emotive and controversial concept. This worries me, 
> > particularly as this random selection of violence-tinged organisations
> > (there are 1000s more) is somewhat weighted towards groups in support of 
> > largely Islamic populations or causes. I would be a lot happier with TAP
> > if you could expose (even just in prose/html) the definitional
> > guidelines you're using to populate these bits of the KB. The single
> > word 'terrorist' does really clarify things.  
> 
> There are in fact a lot more:
> 
>   http://ontap.stanford.edu/terrorists.html

Lots more. Thanks for sharing the new version. I like that the new
version exposes the source of the claims. Very neat. 
I'd still be interested to see a definition (just plain old prose, 
not FOL/OWL etc) for this class.

 
> That download script is broken. The link above goes to newer versions of the
> KB browser, but it's a static page (the last time I posted a link to a 
> development server, it took me three months to stop Yahoo Slurp from crawling
> it despite robots.txt saying to get lost). The source of the data is not 
> currently document in TAP, but it's the State Department's Patterns of Global 
> Terrorism 2002. 

Ah, that explains the slant. They're in the news today btw:
http://www.voanews.com/article.cfm?objectID=83F3C459-47EE-4A24-82943B3856E793F0

[[
U.S. Secretary of State Colin Powell has acknowledged that a recent
government report on global terrorism contains errors and will be
corrected.

The report, called "Patterns of Global Terrorism," concluded that in
2003 the number of terrorism attacks worldwide was at its lowest point
in 34 years. Bush Administration officials had offered the finding as
proof the war on terror was going well.

The finding however was criticized by academics and intelligence
analysts, who said a number of terrorist attacks were omitted, including
attacks in Turkey and Chechnya that claimed several hundred lives.

Mr. Powell denied accusations that the data had been manipulated for
political gain. The secretary said mistakes by a new terrorism data
collection office contributed to an undercount of attacks.
]]

> I think the 7 that show up on tap.stanford.edu were put in during 2001 by
> hand, probably from the same "patterns" source.

Ah, right.  

> [...]
> > My point though isn't to argue the specifics of claims that are in TAP,
> > but to suggest that TAP might see wider adoption if the KB editors tried
> > to structure things so that the claims and worldview embodied in TAP are
> > neutral and uncontroversial wherever possible. I think, even in the case
> > of terrorism, this is achievable, by modeling the controversy (as
> > sketched above). There is no real need for TAP, TAP's editors, or its 
> > publishers (Stanford University?) to say "Organizations X, Y and Z are 
> > 'Terrorist Organizations'", when it could instead say, with less
> > controversy, "Party X denounces Organization Y as terroristic". This
> > has two key benefits. Firstly it makes TAP more widely adoptable by
> > pushing controversy down into reported content, secondly it allows 
> > the basic structure of (certain kinds of) controversy to be modeled and
> > used in applications. Maybe I've said the same thing twice. There are
> > two ideas closely tangled up: being uncontroversial isn't the same thing
> > as being bland, uninformative. Moving to a relational approach to 
> > capturing the information re terrorism would I believe make TAP more
> > informative.
> 
> These sorts of problems are heavily worked on in TAP, the best paper to read
> about it is this one:
>   http://ksl.stanford.edu/KSL_Abstracts/KSL-03-15.html
> It's called "Semantic Issues in Web Scale Knowledge Aggregation" and it
> details many of these sorts of issues, including the two you mention, the
> problem of point of view and the problem of time.

Looks interesting, I think I saw this or a draft before from Guha.
I'll have a closer look...

> A more detailed paper describing first steps toward solving these troubles
> during aggregation is here:
>   http://tap.stanford.edu/contexts.pdf

Ah that was the one I saw. Will re-read.

> The second point you make, about contributions to TAP, I'm not sure is so 
> simple. I'm not convinced that lots of people were considering contributing
> data until they were offended by a simplistic categorization of their favorite
> group or person.

I was thinking at a different granularity. It's more than people like me
who have other RDF vocabs that TAP could fit nicely into might be warier
of endorsing its usage (in tools and software) if it appears to have too
much stuff scraped from dodgy US govt sources. In particular, a lot
of Europeans are particularly sensitive about such things in the last
year or so. TAP's got a reasonably good chance of being the first big 
and cross-domain instance-oriented RDF data source that people start to 
base things around. But the more controversial pieces there are, the 
harder it is to claim TAP as a common base for everyone to build on.


>		 That being said, the Wikipedia model seems to be rather 
> effective in this area, describing contentious issues in a way such that
> controversy is embraced and expressed, and I think more could be done both
> to stay away from the sorts of problems described in the "issues" paper, and
> to stay away from honking people off. 

Yep, Wikipedia seems to be doing pretty well on this front.

BTW re collaborative data production, have you seen MusicBrainz? The 
project seems to be maturing nicely. They have RDF-friendly data model
and Web service interfaces, and CD/Album/etc track listings for a lot of
artists, as well as TRM digital fingerprints of audio recordings.
http://www.musicbrainz.org/

http://www.ldodds.com/blog/archives/000130.html is draft of a schema.
http://www.musicbrainz.org/MM/ outlines the data exposed, with examples
and protocol detail (HTTP-based).

Some data dumps can be found in
ftp://ftp.musicbrainz.org/pub/musicbrainz although I can't connect right
now.

I wonder what the overlap is between TAP and MB... 	

Dan

Received on Friday, 11 June 2004 21:09:36 UTC