I've been pondering the current push for standardised terminology in 
Museum classification so that different Museums' IT systems can be 
interfaced, and I'm wondering how much of this is really thought-through 
... Our vision for the "future" of Museum IT seems to be based onthe 
needs of ecommerce systems, or on how information technology was taught 
in the 1970's.

Museums don't usually need to be able to bulk-exchange data with each 
other, because they're not wholesaling or retailing each others 
exhibits, or buying and selling in bulk from each other and expecting 
their inventory to be automatically updated. They're not Amazon or 
iTunes. The reason for making information available online is usually 
for the benefit of end-users, and those end-users are primarily 
interested in finding out about exhibits and finding other similar 
information. They don't need to add or remove exhibits from the system, 
or transfer entries to a different museum's site. If they're not moving 
entries between systems, then deep data-compatibility isn't really a 
requirement. If your museum is going to be taken over or merged with 
another, then it's handy if the organisation that takes over your 
collection can integrate your database with theirs, but otherwise, it's 
a bit difficult to see the immediate payoff.


If what we want is search and discovery, then structured XML databases 
become less critical, and the specialist dedicated tools for the job are 
... search engines. Good search engines look for patterns in the data 
and find their own sets of associated keywords and cross-references 
without needing webpage authors to standardise on a specific keyphrase. 
If you search for "747 aeroplane", Google will report Wikipedia's page 
on the Boeing 747 as the highest ranking result, even though one of the 
two selected words, "aeroplane" doesn't actually exist anywhere in the 
page. Google knows from context that different pages that include "747" 
seem to use "aeroplane", "aircraft" and "airplane" in the same way, and 
it makes the association that, in this type of search, the words are 
probably interchangeable. Google also doesn't need those keywords to be 
explicitly structured in the source text (although it probably helps). 
Google also has access to semantic structures via sites like Freebase 
that can tell it that "747" is a type of "airplane" / "aircraft" / 
"aeroplane", which is a type of "transport", and which is associated 
with a "manufacturer" called "Boeing", so it can draw on these logical 
associations and use them to guess at the meaning of museum webpages 
without needing those pages to include their own semantic tagging.
Explicit semantic tagging probably /helps/, but one of the points of 
teaching Google about semantics separately was that it could then use 
that knowledge to analyse /any/ webpage, instead of requiring thousands 
of individual webpage authors to go off and take special training 
courses in standardised terminology, to be able to write pages that 
Google can understand. That seems to be the equivalent of what we're 
asking museum staff to do, with the added downside that once they put 
all the effort into structuring their data to allow it to be more 
compatible with some hypothetical inter-museum system ... they find that 
no such system seems to exist. I'm not even sure that anyone's even 
planning on producing one, or setting up the organisation to run one, or 
sorting out what the rules would be if one existed.

So, if our supposed goal is to let people cross-reference and search for 
similar items across the Museum network, perhaps what we should have 
been concentrating on is a cross-site "Museum search" project. The 
end-user doesn't need any of those museum computers to be talking to 
each other, as long as there's a search engine button or widget that can 
be launched from any of the sites that gives access to pooled results 
for all of them.
As far as I can tell, the reason why we haven't done this is because the 
Museum IT community has been focusing on XML as the exclusive answer to 
everything, because XML is nice and technical, it lets them impose 
order, and it requires IT people to understand it so it makes Museums 
more dependent on IT people (which IT people probably feel is a Good 
Thing). XML-based initiatives generate IT jobs, and IT training jobs, 
and IT support jobs, and if you can lobby the standards committees and 
get your XML based system or scheme made compulsory as a condition for 
certification, then museums have to keep paying you, indefinitely. 
They're locked in, even if your complex system doesn't actually do 
anything especially useful.

So perhaps the problem with a search engine initiative is that it might 
work /too/ well. It might be too quick, easy, cheap, effective and 
popular. If the search engine functionality is being implemented at a 
single point, you don't need specially-trained IT staff duplicated at 
every single museum entering data in a special way that the IT system 
requires. Sure, if you /want/ to use explicit XML tagging, that might 
give you a boost in the search engine rankings because the engine will 
have a higher confidence in its analysis of a page, but if you just 
write a simple webpage about an exhibit, and tell the dedicated Museum 
search engine that it exists, then there's a good chance that the engine 
will be able to do a good speculative cross-reference without needing a 
single line of custom code.


---------------

One way of implementing this would be to have an "Exhibits" widget that 
a webmaster could embed on any page that's about a single Museum 
exhibit, which would then register that page with the search engine when 
it's loaded, and give the user "Search for similar items on this Museum" 
and "Search for similar items in other museums" options. Maybe also an 
"I like this exhibit" button, a button to look for the current ranked 
favourites in the site, and a star rating based on how well that exhibit 
is ranked on the site.


From the Museum's point of view, this wouldn't seem to have to be any 
more difficult than embedding an existing social media widget, and for 
many museums it might work well enough with existing content to make 
more ambitious semantic tagging projects unnecessary. If someone's 
looking at a page with "Steiff" and "teddy bear" in a heading, a 
dedicated "Museum search" for pages with similar content probably 
doesn't need those keywords to be semantically tagged to be able to find 
other Steiff bear exhibits. Additional structure would be nice, but 
usually unnecessary.

If you want to get more fancy, you could have a Class="Exhibit" 
identifier that could be put into the enclosing div or table code, to 
say that only the contents of that particular panel are relevant, so 
that the search engine doesn't try to index all the surrounding 
navigation bars etc. If you wanted multiple exhibits on a page, they 
could have their own widgets and isolating panels. But that could be a 
later development if people wanted it.

I do like the idea of having everything XML-tagged on principle, and I 
think it's a good goal to aim for. But if we're serious about wanting to 
let users do cross-museum searches, XML seems to be the foundation work 
for a very sophisticated house that nobody's intending to build. If we 
honestly do want cross-museum searches, we can have it without a lot of 
work, but the limiting factor is people, not technology.


OTOH, if we actually don't care too much about the ability to do 
cross-museum searches, feel that maybe they won't be all that useful, 
and aren't too bothered if the feature never appears, then that's okay 
... as long as we're honest with ourselves about it.
Eric

****************************************************************
       website:  http://museumscomputergroup.org.uk/
       Twitter:  http://www.twitter.com/ukmcg
      Facebook:  http://www.facebook.com/museumscomputergroup
 [un]subscribe:  http://museumscomputergroup.org.uk/email-list/
****************************************************************