|
Users of scientific handbooks often have a
fairly specific information need: if this is the case
they want to avoid having to read or scroll dozens of pages, they rather need a
way to ``jump'' to specific excerpts of the handbook covering the topic
they are looking for.
In other cases, users of handbooks have vaguer information needs,
in which case the ability to produce a query may be hampered by the poor
knowledge of the domain.
These considerations made us think of an environment where the user
cannot only type in queries, but can also take advantage of an
organization of the material that can meet her vague information needs.
In reference works, such as dictionaries and encyclopedias, the subject is usually arranged in
alphabetical order so that topics may be located quickly and easily.
In books and handbooks, the subject is organized into
chapters, and usually an index is provided to serve as the direct guide to the
many topics treated in it, or to locate the smaller
subdivisions of the larger subjects. The
back-of-the-book index may also be organized in several disjoint
parts, such subject index, author
index, name index, etc. Also, another important element in the arrangement of material in a
reference work is the cross-reference that will refer the
reader to additional related information.
From the back-of-the-book index comes then the inspiration to organize the
subject in a way that is informative for users with vague information
needs and possibly poor background in the area. Therefore we thought
of a map of the domain, that beyond continaing references to the
appropriate places in the text, also includes relations between the
elements of the map, so that the user is allowed to ``navigate'' the
map also independently of the text.
To make matters more concrete and tangible, we
work with the domain at the interface
of logic and linguistics. In this domain we identify important terms,
and relationships between those terms; these will be used to build a
browsable map, which is then presented as the reader interface for the
handbook.
Given this approach, the research questions we address in this thesis are the following:
Question 1
What requirements should we impose on a map that is to be used for human browsing
and as a skeleton to provide focussed access to the text? And how
is the application of the requirements going to work?
In the case of the Handbook of Logic and Language, Research
Question 1 comes with an important constraint:
the process of populating browsable map should be as much as possible
a collaborative and bottom-up process.
There are several reasons for this constraint, some principled, some
pragmatic. The interface of logic and language is evolving rapidly,
and highly interdisciplinary. Moreover, unlike, say, medicine or law,
there are no resources (or standard bodies) to support imposing a
standard in a top-down manner: (expert) colleagues around the world,
and from around the interdisciplinary area covered by the
Handbook of Logic and Language, will be the ones populating the
browsable map with terms and relations between terms.
Question 2
How do we present the map to readers of a handbook in such a way that we ensure broad
coverage of the domain (with detailed information per term), while
making sure that users do not get lost?
The answer to Research Question 2 will be an
user interface able to accomodate detailed pieces of information
about the topic in the map, while at the same time giving the user
ease of navigation and good feeling for the "general picture."
The interface should also able to equally support searching and browsing,
in order to allow users to satisfy a broad variety of
information needs and user background, from the more specific (search) to the
more vague (browsing).
Assuming that we are able to come up with a satisfactory answer to
Research Question 2, we need to hook up the
browsable map to the documents of which it is meant to provide a map.
There are two aspects to this, covered by the following questions:
Question 3
What are suitable targets for focused links from terms in our
browsable map to documents in the handbook? And how do we
identify such targets for a given term in our browsable map?
Notice that in Research Question 3 we ask for
focused links from terms in the browsable map into the
handbook, and not just for links from a concept to an entire chapter
of dozens of pages. This is in line with our wish to provide focused
access.
An obvious key issue along the way will be evaluation: how do we assess our
proposals? At different stages of our work, different types of
evaluation are appropriate.
When we introduce our proposal for a
browsable map that generalizes the concept of the back-of-the-book
index to the setting of electronic scientific handbooks, we compare
the results we obtained with the requirementes we set ourselves, but
also with the internal organization of the handbook.
Later, we perform user studies to assess the effectiveness of our
proposed visualization method for exploring the map.
And a third type of evaluation is conducted when we link the terms in
our browsable map to text segments in the handbook; this is the type
of evaluation that one encounters in information rertieval and applied
language technology, where we develop "gold standard" corpora of
ideal outcomes and measure the performance of algorithms against this
yardstick. Since no standard test sets exist for the issues that we
are tackling, we developed our own.
Question 4
In the "old" days, authors and editors were often the ones that
had to provide the entries for the back-of-the-book index. In our
proposed set-up, how do their roles change? Will they be expected
to populate the envisage browsable maps? Will they write
differently, knowing that a map will be linked to their texts?
| 