Re: Hyper-G and MARTIF

Gerbert Orasche (
Tue, 06 Feb 1996 11:25:02 +0100


At 22:39 05.02.96 GMT, you wrote:
>On Mon, 5 Feb 1996, Lesley West wrote:
>> In article:
>> <>
>> writes:
>> I suggest you clean out anything not specifically related to protein
>> structure, such as enzymes. You'll find the existing links already
>> groups implicitly.
>> Then start grouping and splitting. In my experience this is pretty
>> contraversial as people fall into splitters or groupers thus leading
>Well said , lesley. This is an area where everyone will have ideas an
>there is no 'right' way of doing it. It's quite easy in such cases to
>get into religious, or flame- wars :-) So we *must* remember that 'my'
>view is not the only one, although it certainly deserves an airing.
I have to say in the beginning, that my knowledge of chemistry is that of
highscool and thats quite a time ago.

But I see your problem. Different people have different approaches to a
glossary. This problem does not come up only in your special case but
everywhere in huge (hypermedia-) databases. At developing Hyper-G we were
discussing how to solve this. Our solution was to provide additional
hierarchy information, where objects (we denote as an object all types of
documents like HTML, PostScript, Images, VRML scenes, but also link anchors)
can be replicated in the hierarchy without being stored in the database more
than one time.

What does this mean to you? You can provide as many approaches (sub-trees)
as you like and have to edit the data objects only once and put them into a
main collection tree (Hyper-G term for a directory-like object in the
database). The next step is to build up any hierarchy (splitting, grouping)
you or others like to see. The user can now select the way to browse the
database he likes best.

The clue here is, that data managemant effort for the provider is minimized.
If you want to change one object (in your case e.g. a specific molecule) you
edit it once and get the actual version in every subtree. Also if you add
one object it also shows up in the other (directly linked) collections.

>> hours of discussion. In the end, pragmatism about how you search and
>> your search tools must drive your groupings.
>Yes. I hope that Hyper-G may help here. Let's assume that we have a
>term that fits into 2 concepts (both of hierarchival nature). Examples
>8could* be : enzymes and structural families. So TIM is an isomerae and
>a TIM-barrel. (In that case we should produce them independetly and try
>to crosslink)
Here the next powerful Hyper-G feature comes into play: The search facility.
We support several search areas:

- whole server
- selected subtree(s)
- multiple server
- all (Hyper-G) servers in the world

You can then do a fulltext, title, or a keyword search. For you that means,
that your users can define to search only in their preferred subtree (to get
exactly that molecule they want) or whole server. The latter provides not
only the object itself, but also where it is located in the collection
hiearchy. So you can see some kind of meta-information procided by the
creator of the glossary.

These are only two of very many features and I really think that your work
will be much easier to edit and better to use, if you use Hyper-G.

Please feel free to ask questions, if I was not clear enough.

Gerbert Orasche at Graz University Of Technology
IICM (Institute for Information Processing and Computer Supported New Media)

Tel-Net ++43-316-873-5607