[Sigia-l] RE: Navigating through three dimensions

[Matthew C. Clarke]

javier velasco writes:

> I have a huge archive of articles, mainly news articles. And my plan is
> to organize a browsable archive to engourage navigation, wandering. My
> approach is to organize the archive into tree axis:
>
> - Subject - remember "my first taxonomy"? it covers sports, culture,
> society, education, etc. very huge.
> - Publication (Media) - 4 newspapers, 7 magazines, some other media and
> their corresponding sections.
> - Date - publication date.
>
> I believe these three axis sytem qualifies as a faceted system. Am I
> right?
>
> So where I need your help is in taking it to the interface level, please
> send me examples of interfaces that manage this kind of stuff, or ideas
> on how I could keep on moving.

It seems to me that there could be some profitable cross-fertilization
between complex text search spaces and multi-dimensional database (MDDB)
techniques. An MDDB (often called a cube) is often used to store a data
warehouse, i.e. a consolidation of data from other (typically relational)
databases. The dimensions for a sales cube may be something like time,
product, customer age, customer gender, branch at which sale occurred etc.
And at each point in that multi-dimensional space the cube stores data items
such as sale price, cost price and profit. The cube has aggregated values
pre-calculated so that a query such as "Find the total profit for all
widgets sold at Californian branches during 2001 (regardless of customer age
or gender)" can be answered very quickly.

A typical user interface for navigating a cube is a pivot table. That is, a
2D display where one or more dimensions from the cube are nested on both the
horizontal and vertical dimensions of the table. (Excel can do things like
this if you want to experiment with the idea.) You can drag-n-drop
dimensions onto the table and filter the dimensions by selected values or
groups of values.

e.g. a sales pivot table may show customer age groups down the left-hand
side, and for each age group shows a row for Male and another for Female. At
the same time, the time dimension is shown along the top row, perhaps
grouped by month, and a column for each sales region. The user can easily
specify which of the data values should appear in the table's cells, or
change the grouping from month to quarter, or adjust a filter so that only
months in 2001 are shown, or add/remove a dimension from the left, or move
the gender dimension fro the left to the top etc.

So much for the background info in case you haven't played with MDDBs. Now
for my point.  If you have a multi-dimensional classification system (why
stop at three?), perhaps you could implement an approach to navigation akin
to a pivot table. Provide a UI that allows the user to configure the
dimensions *they* want to manipulate -- maybe putting Subject down the left
and Publication along the top, with columns for each year within the
Publication categories -- and mechanisms for filtering those dimensions.

Now the problem that I'll leave for homework is what to put in the cells of
such a pivot table. In an MDDB situation it is a meaningful number, but what
it could be in a text-space situation will depend on the exact context.
Could be the number of items that fall into the location of the
search-space, which can be clicked to show the list of those items. Could be
a hyperlinked list of article titles, or authors etc.

I'm not suggesting this is the solution for your case, just acting like a
bee doing a bit of cross-pollination.

Matt.