[UCR] Strawpoll - Information Integration

[The strawpoll process requires a separate email rather than an inline
comment when trying to indicate that we feel a section is not yet ready. So
here's the associated email for the Information Integration section.
Seems like a rather heavy weight process for just sounding out WG feeling.]

When read as a standalone description, rather than in conjunction with the
Wiki pages, this section does not make the role of RIF in information
integration as clear as it could be.

I think what is needed is largely a tightening of the language and
structure. However, I was sufficiently confused about parts of what was
being said to find it hard to simply suggest modified text.

The mention of "two primary integration techniques" suggests that this
section is about integration of information that is *expressed in* rules,
rather than about integration of information by *use of* rules. If that's
the intention please state that more clearly. If it's not, then this
paragraph needs more context.

Scenario 1: Job seeking
This scenario is not very compelling as stated. It's something of a leap to
suggest that employers might publish adverts which include rule-based
mappings from local to shared ontologies rather than publish in the shared
ontologies in the first place. The suggestion is that this is needed
because the employers need to use concepts not present in the shared
ontologies yet an extension ontology making use of subClassing (or
thesaurus style relations expressed via SKOS) would seem more appropriate
than rules for this. If that's truly the scenario then a concrete example
might help to make the problem, and way RIF helps, clearer.

Scenario 2: Product launch planning
This talks about pulling RDF data from multiple internal databases and
integrating it along with supply-chain contract descriptions represented as
"RuleML/OWL/Prolog facts". All of that information appears to be
collections of facts, not rules. The role of rules is not clear from the
description. As stated its seems to be a good use case for SPARQL (protocol
as well as query language).

Minor comments:

o The list of sources of heterogeneity should either be linked to some
mention of whether or how RIF helps address them, or be dropped.

o It might help to clearly state early on that there are two separate,
unrelated scenarios to be described.

o It is unclear whether para 5 ("Company information ...") is supposed to
be part of one or other of the scenarios. It's a true comment but seems
orphaned where it is.

Dave

Received on Monday, 20 February 2006 15:29:58 UTC