background image
DigiCULT 15
adaptive interfaces - were developing fast but, he said,
the heritage sector was not leading its own, unique
researches, just following.
Forum Moderator, Joemon M. Jose, a multi-
media information retrieval expert with the Com-
puting Science Department at the University of
Glasgow, Scotland, thought the experts' approach
should be from three viewpoints: e-learning, `virtu-
al museums' and `presenter', which he described as `a
virtual artist who wants to create a multimedia pres-
entation or a documentary creator who wants to
identify material for his project'. He went on: `So
the first question is: which available and potential
resource discovery mechanisms are most relevant for
heritage institutions?'
E
VALUATING
THE
FOUND
OBJECTS
B
ut, Swedish Institute of Computer Science
researcher, Jussi Karlgren, put his finger on a
concern that needed solution before they did that: in
other words, identifying and verifying the sources of
information gathered by these mechanisms. All three
of the Moderator's scenarios could be challenged to
evaluate material assembled and presented by an auto-
mated system, he suggested.
`If, for example, I wanted an overview of Finn-
ish history and found a Russian museum, a Finnish
museum and a Swedish museum to give it, the three
pictures would be completely different and so con-
tradictory as to challenge the person receiving them,
even cause distress. So how do you provide these
intellectual challenges in ways that are obvious and
yet get answers you can believe.'
Worse could happen, said Dr Karlgren. `I can imag-
ine scenarios where a highly pres-
tigious institution would say "I will
not be a part of it if that institution
is part of it and if our informa-
tion is mixed", which is complete-
ly likely.'
His countryman, Lund Univer-
sity librarian Koch, had an answer,
though not an immediate one:
`This calls for provenance meta-
data to a much larger degree than
we find today. There are lots of
text and documents on the Web
that you won't be able to identi-
fy, sometimes not even the organi-
sation from which they emanated.
You would not be able to identi-
fy who is the intellectual produc-
er or owner of the information, statement, opinion, or
text passage.'
He went on in words aimed at institution cura-
tors: `Because of these different views, there are much
higher requirements for even small chunks of text
passages and other stuff to be identified with the cre-
ator, the creator's affiliation and, maybe, point of view.
That could not be neutral, of course, but it would lat-
er on allow text criticism and analysis of prevailing
cultural and political views.'
Moderator Jose wondered if information conflicts
might be detected by the search system, an idea that
appealed to Dr Karlgren. He teased: `Detecting con-
flicts automatically sounds like a really wonderful
research project I would like to work on for the
next ten years or so.'
However, he thought that users would be aware
that conflicts were likely even if the information was
authentic, authorised and reputable. He suggested: `It
is quite likely that Vikings are viewed differently in
the north of France than they are in Scandinavia, for
instance.'
He went on: `Objective judgement is difficult on
the exploration of northern Europe by seafarers or,
alternatively, the rape and pillage of the northern seas
by barbarians, unless you know that this information
is from this source and this is from that; people trust
this information source; this bit of information is used
in many places and those in turn are reputable places.'
The way to deal with the problem was the use of
provenance to identify sources and reveal how oth-
er users viewed their reputations. Then it would be a
question of trust, in much the same way as one chose
a newspaper. But, he wondered, how would such
metadata evolve?
Dr Andreas Rauber had a solution. He is an
associate professor in the Department of Software
Technology and Interactive Systems at the Vien-
na University of Technology in Austria, a student
of machine-learning techniques to make collec-
tions more accessible. Instead of assigning metadata to
objects, the objects should be associated with infor-
mation, he proposed.
He explained: `Put the object into a kind of ... well,
I wouldn't like to call it an ontology because that's
way too structured, but put it into a kind of informa-
tion space. Just as you have information about knowl-
edge, so you have information about certain periods.
Put the object into that and then describe the sit-
uation, the relationships. As soon as you add that
description it becomes available to all the objects that
are placed in that period.' `You have different views
of certain events then you place the object into them.