Hi,

This is an interesting area - I think it points ahead to the future uses of
object and collections information and I'd be interested in any input.

At the EVA London 2003 conference Michael Lesk (http://lesk.com/mlesk/ )
argued that we are reaching the useful limits of classifying and ordering
information: the way ahead is to develop better and more sophisticated ways
of searching (which is one of the aspects of web services, in fact). He
illustrated his point by comparing the results from an academic portal with
those from the usual search engine - guess which was obviously most
comprehensive, and not that hard to select useful sites, either.

I later came across some work in the San Francisco Art Museum - a concept
called Word Soup - where descriptive keywords had been ascribed to the
paintings by volunteers (with some training), using terms they thought
described it rather than academic art history or curatorial terms, as these
are the terms that ordinary people are likely to find useful. The curators
had not liked the idea so the collection of terms was called the Word Soup.
http://www.4d.com/solutions/thinker.html and http://www.thinker.org/.

These ideas are clearly all related and are pretty much what John is saying
below, as well. Think "search" rather than "classify", and you may find ways
of enabling users to input useful terms to search on. Of course one way to
do it is to capture the terms they are searching on and work with those ...
The difficulty at this stage may be to automate the capture of user
generated terms or information.

This is probably all related to the Semantic Web too, I expect?

I'd be interested in further discussion on this, off group if it gets too
detailed,

Suzanne


on 8/9/04 9:30 am, John Faithfull at [log in to unmask] wrote:

> Hello,
>
> Some quick personal thoughts...
>
> Most (all?) attempts at structured "whole world" classification fail because
> it's very hard - esssentially this is what human languages do, and there is
> no "right" answer. Languages are easy to learn, hard or impossible to model
> effectively, or implement synthetically as an end-user tool.
>
> I think attempts to rigorously structure object information for all possible
> purposes (other than for a discrete use, such as collections managment, or
> conventional specialist purposes) will always fail. The most fruitful
> approaches approaches involve the use of descriptive or classificatory
> keywords, or interpretive text in conjunction with free-text searching.
> Application of such keywords, phrases or text would be subjective, but this
> is unavoidable - you cannot entirely formalise a subjective process.
>
> A. You could decide to use a restricted set of such terms - in which case
> the structure can be imposed not on the objects or their descriptor
> categories, but on the keywords themselves in a kind of mini-thesaurus.
> Because your thesaurus doesn't have to include all terms about everything,
> it's much easier to set up and to make usable, and indeed explain to users.
>
> B. Or you could use open keywords - augmenting the record with any
> descriptive or classificatory words or phrases you think might be
> appropriate. This has the advantage or richness, and probably similarity to
> normal human language (and hence user expectation), but it is difficult to
> classify all such terms meaningfully. This approach is also future proof -
> it is not constrained by your current idea of what you want to do, and how
> you want to do it.
>
> C. Or you can use both approaches A and B together - not actually difficult,
> and probably the most fruitful approach.
>
> The other problem is how to imbue the words themselves with context. One
> approach might be simply to apply these to traditional "areas" of an object
> record eg what, who, when, where. This is simple, and can work pretty well.
> More is much harder, and is subject to diminishing returns. If the
> application of descriptive terms to records is subjective, equally so is
> application of context to these term: there is no definitive answer to this.
>
> For example an object associated with the battle of Waterloo, could have
> some sort of
>
> place: waterloo
> type of event: battle
>
> information (even this example throws up all sorts of logical problems!)
>
> but nobody actually thinks like that, and analysing the whole rich world of
> all objects and contexts in such a way is impossible - that's why we have
> language and grammar. You get round all these problems by simply using the
> human language phrase:
>
> battle of waterloo
>
>
> To sum up, I'd recommend
>
> (1) Develop a small restricted keyword vocabulary designed to provide a
> pointer to general areas. It may also be possible to combine this with a
> small number of context keywords, or structure. Do not attempt to encompass
> everything. You will fail. Limited aims are achievable, and will help a
> great deal in guiding users.
>
> (2) Encourage the recording of "subject matter" keyword and phrase
> descriptors about all types of object and don't worry too much about
> terminology control for this kind of description. Use human language, and be
> rich and helpful with your terms.
>
> Together these will enormously increase the accessibility of your data.
>
> Better leave this now...
>
> Cheers
>
> John
>
>
>
> -----Original Message-----
> From: Museums Computer Group [mailto:[log in to unmask]]On Behalf Of H
> Hollis, Archaeology
> Sent: 07 September 2004 21:41
> To: [log in to unmask]
> Subject: Artefact interpretation online
>
>
> Greetings everyone. (Apologies for cross posting.)
>
> I am working on a digital database structure that would allow users to help
> structure the categories by which archaeological artefacts (including
> objects, documents, video footage, oral history) are classified within the
> database.
>
> For example, a photograph from Kenya in the 1930s would be classified by
> time and place, but might also be classified by the content of the
> photograph (elephant tusks) or activities associated with it (trade: ivory)
> or indeed by how the artefact has been used (to teach KS3 History about
> Empire, or in a brochure for a museum exhibition). Those who use the
> database build up the meanings for that artefact over time.
>
> Does anyone know of similar work already in place on the web, for whatever
> size community base? Often I find that digital archives/databases for
> museums narrowly limit the artefact into very particular categories -- I am
> interested in a more multi-vocal interpretation of the artefacts.
>
> Any input no matter how tenuous is welcome.
>
> H Hollis
>
> ----------------------
> H Hollis, Archaeology
> [log in to unmask]



///////////////////////////////////////////

Suzanne Keene
162 Erlanger Road
London SE14 5TJ

Telephone: 020 7639 5371
mobile: 0779 962 7002

///////////////////////////////////////////