ALEG

Introduction

This document presents some ideas on how a topic map based structure could be browsed by an end-user. It isn't meant to define a proposed interface - just explore the benefits and drawbacks of one simple option based on the Windows File Explorer navigation paradigm.

In particular:

• Such a browse facility would not replace a search facility. Users wanting to find something in particular would probably search rather than browse.

• Windows File Explorer is not a particular good UI, even for browsing homogeneous contents such as drives, directories and files - I'm always losing context in a deep directory display! Perhaps something to highlight my current position in the hierarachy (bold lines, bolder text showing the current 'path') would help. But, I'm using it here because it is easy to 'mock up' a display, as shown by Rafal Ksiezyk's XML Europe 99 paper, Trying not to get lost with a topic map.

• Other visualisations are possible. For example take a look at an example of using Topic Maps to link concepts in a play - this is a smaller and simplier problem than the one we must solve with ALEG, but interesting nonetheless (note - drop-down selection lists don't work with other than tiny databases!)

  Other approaches include:

  • Wobbling, gyrating lines and boxes. The Brain - a dynamic map which changes to show the current focus and immediate nodes in the centre of the display. A simple demo is available somewhere on ZDNet (I found it yesterday, but I searched The Brain for it today in vain...). What do you think? This has been around a while (in web terms), but hasn't taken over the world by any means. I find the context shown a bit limiting, although it is much easier to use than the Smithsoneon Revealing things experiment which takes a similar but more ambitious approach.

  • TheBrain-like:
    • Lucid Fried Eggs - topic like with 'notes' having multiple parents, children and 'jump' relationships with other 'notes'
    • www.memes.net - an application of Lucid Fried Eggs - pretty cool!

  • The Law Department web site has been built using topic maps. It takes a conservative, simple approach which works well within its domain.

  • Inxight's 'Site Lens' (Barnes and Noble example) demo and 'Hyperbolic Tree' (HR department) demo. And a Hyperbolic Tree 'walk-thru demo'.

  • Cartia's demo of NewsMaps plots topics in a make-believe geography. Try the "today in history" link on the NewsMaps page.

Top Level

Fig 1

Region

If we expand the Region topic, we might find something like this:

Fig 2

1. All topics which are subdivisions (aka subclasses or specialisations or is a) of this topic (Region)
2. All topics for which there exists at least one relationship entry in the database which nominates this topic (Region) as a participant in the relationship

Pedantically (and economically), because the is a relationship of a subclass is represented as a relationship between the subclass its "parent" class, this could be stated using just the second condition, as the first condition is a subset of the second.

So, under Region we have the above listed subclasses of Region, but no actual works or creators because none of these have been linked to this rather abstract notion of Region. So far, so good....

Expanding Region's subclasses

If we expand each of the Region subclass topics, we might find something like this:

Fig 3

Expanding Sydney

If we expand the Sydney topic, we might find something like this:

Fig 4

What we are showing here is very different from what we've seen previously! In the previous examples, the topics we'd found had all been related to the topic we were expanding with a "is a" relationship:

• city is a region
• physical region is a region
• social region (should that be socio-ecomonic? take all these names with a grain of salt!) is a region
• state is a region
• town is a region

• Melbourne is a city
• Sydney is a city
• Blue Mountains is a physical region

But here, none of people, suburb or work have an is a relationship with Sydney. Instead, the scan of the database to find topics in a relationship with the Sydney topic has found many topics and has grouped what it found by showing the relationships of those topics to the Sydney topic.

Why do this? Why not just show all the topics? I'm not sure! Certainly, we could. But maybe there would be so many topics of such different types that the user would be confused/swamped with information. It is obviously important to show the relationship of the found topics to the base topic - "Born in", "Subject Of" and "suburb" are completely different relationships, and it is unlikely that a user would be equally interested in all three! So, the visualisation could group them by the relationship, and represent the relationship with:

• a name (as shown here),
• an icon (tricky), or
• both (maybe optimal for frequent users who learn to assoicate the icon with a relationship type and can hence avoid reading all the words!).

It could show:

• just the groups (as shown here)
• the groups and their contents (other topics) in "sub folders"
• the groups and their contents (other topics) but not in "sub folders"

As the user is unlikely to be interested in all the groups, maybe "expanding" them pre-emptively is the wrong approach; better to let the user look and decide what they want next.

Expanding born In

If we expand just the born In topic, we might find something like this:

Fig 5

Finally! Pay Dirt! Here, we've scanned the database to find topics for which there exists at least one relationship entry in the database which nominates this Sydney topic as a participant in the relationship and with a relationship type of born In, and we find a couple of people. If the user was looking for Australian Literary figures "born In" Sydney, maybe now they are happier!

But maybe not... There could be hundreds or even thousands listed (in what order? Date of birth order? Surname? Citation or relevance ranked?). Some may appear to have the same name (Smith, John...).

The order and name problem are easily solved - maybe a "right hand window pane" opens (just as is displayed by the Windows File Manager) which lists the entries in more detail - maybe showing in columns:

• date of birth and death
• gender
• pseudonym information
• ...

The user could sort by clicking the appropriate column header (as in Windows File Manager - a familiar UI technique). Maybe the user could even filter the list, using techniques found in Excel, although the knowledge we are expecting of the user is getting quite arcane at this stage!

Alternatively, we could group the people "born In" Sydney by some technique such as:

• Alphabetical groups: maybe A-D, E-H, I-M etc. How many groups we show, or even whether we show any could be dependent on the number of topics (people in this case) to be shown.
• Some more complex attribute of the topic. For example, group by gender, group by date-of-birth range (1700-1880, 1881-1920, 1921-1955, 1956-), group by genre. Probably a case could be made for these and more! Maybe the user could set default groupings by expressing a preference for seeing things grouped in a time basis (but I fear we are just twining rope for the user's neck).

Expanding John Bundaberg

Anyway, pressing on, if we expand the John Bundaberg entry, we might see something like this:

Fig 6

Here we are showing a mixed bag of topic groupings:

• Relationships which groups all the topics which nominate John Bundaberg as one end in the topic relationship (confusing, sorry) which concern "big R" relationships - familial, influenced by/ influence on, maybe relationships with organisations/movements...
• Works about which groups all topics which have as a subject-type topic "John Bundaberg" (and what about John Bundaberg's pseudoymns?)
• Works by which groups all topics which have John Bundaberg in some sort of creator role (author, editor, illustrator, ...)

Other groupings you'd expect to find in a more comprehensive example would include:

• pseudonyms of (hmmm... would the pseudonym itself appear as a 'top level' entry under 'born in Sydney' if relevant to the pseduonym's persona?)
• vocations associated with John Bundaberg
• time periods associated with John Bundaberg (maybe he had a "depression era" bent, and the "Depression era" topic was deemed worthy of linking to him
• subjects tackled by John Bundaberg's work, eg "poverty"

Note: Just what would topics such as "depression era" expand to? I don't know. What about "poverty"? Here at least it would seem reasonable that "poverty" wouldn't expand to show subject narrower terms (if it were a thesaurus word), nor would it expand to show all the works with "poverty" as a subject, but that it would just show John Bundaberg's works with "poverty" as a subject.

And what about if we were looking at John Bundaberg not because he was born in Sydney, but because he wrote about Sydney? Then, under the Subject grouping would we only show subjects that co-appeared in works which had Sydney as the subject? This is an interesting approach to incremental searching:

• show me writers who works about Sydney
• thanks, now lets look at them..hmm, this John Bundaberg - lets look at his subjects - oh, here is one on "poverty" - if I expand that, do I expect to see John's works on Sydney and poverty, or will I see all his works on poverty?

If we expand all these groups fully we might see something like this:

Fig 7

Here we see 2 relationship types (influences and personal) which may further expand to subtypes if warrented. We also see "works by" expanded to show:

• Creative works, Essays, Reviews (again, the names/topics may be inappropriate)
• Creative works expands to his 5 creative titles
• One of his creative titles, "Cronulla Short Stories" is furher broken down into these groups:
  • Components Of which reveals the component works of this collection
  • Creators which would show creators responsible for the work
  • Instantiations Of which would lead to information about editions
  • Reviews Of which would lead to reviews of this work
  Maybe manye works would have a "subject of" grouping as well, leading to works about this work. Maybe reviews are part of this grouping, maybe not!

What about holdings? Yes, I guess that might be represented here at the work and/or instantiation level.

What is missing of course is the "beef" - the full biographical details and the full bibliographic information about the work. How should this be shown? Possibilities include:

• In the right-hand window pane?
• In a popup window?