Lists Home |
Date Index |
Mike Champion wrote:
> --- Bill_de_hÓra <email@example.com> wrote:
>> And unfairly, I could twist your
>>argument as being
>>equally against relational data, though I'm sure
>>that's not your intention :)
> Uhh, not my intention, but the comparison is apt.
> Relational data assumes that field values are from a
> domain of well-defined types, and a well-defined type
> is something very closely related to an ontology,
> AFAIK. RDF-ish ontology / inference systems can model
> semantic networks in a more natural (to ordinary
> folks) way than relational normalization and joining,
> but that's an implementation detail :-) So, I don't
> see much *conceptual* difference between "improve
> search by building ontologies" and "improve search by
> modelling all your concepts in relations", although I
> presume the semantic web will be more web-friendly!
Sure. I think semweb folks would argue that graphs are a more
flexible/decentralised approach than relational databases - for
starters you'd think it'd be easier to merge two RDF graphs than two
db schemas. Fwiw, I think a lot of this semweb/ont stuff will
initially get traction behind the firewall.
> Sure, I agree. So long as one is talking about using
> a relatively small amount of hand-generated metadata
> to make inferences about, or remove or resort the mass
> of autmatically indexed data that a search engine
> uses, I have no quarrel. I just don't have much faith
> in the idea that that ontologies or hand-authored
> metadata in general can do the brunt of the work in
> searching the web.
I'd agree in general - 'low-tech' approaches like foaf/geourl can
help in specialized areas (such as blogs). A key issue with this
stuff will the same as it was in the 1980s, keeping down the cost of
entering quality data.
Bill de hÓra