Re: Hyper-G and MARTIF

peter Murray-rust (
Mon, 5 Feb 1996 10:11:33 +0000 (GMT)

Kurt (and other readers)

Sorry if I have been obscure or ambitious on this - I'll explain... :-)

On Mon, 5 Feb 1996, Kurt Giles wrote:

Glossary construction can be quite involved when done properly because we
have to think of the relationship of all the terms to each other.
Although this isn't essential - it is *possible* to have a very useful
glossary with a simple linear organisation - it helps to structure the
relationships. The MARTIF approach supports this structuring.

In PPS96 which will be a showcase :-) we need to tidy up some of the
structure from last time. There are terms which shouldn't really be in
and otehrs which fall into categories (the word concept is preferred in
terminology, I think). the advantage of this is that we can then locate
all the terms in a given category and make their presentation consistent
(e.g. all the amino acids should have connections table,s , point to
Klotho, etc It should also be possible to find out all the amino acids
at one time.

> I don't think the 'subtree' idea will really work since we would inevitably end
> up with an 'OTHER' where we put in all the things we cannot fit elsewhere, and
> I think this is a bit messy.

Organising glossaries isn't simple! the 'subtree - the concept/category
works well for many systems (sorry about the typos - very bad line). So
it is useful to see what *natural* groupings there are already in PPS95.
If the right concepts are added then there are many ways that the info
can be displayed or used. (It's always possible to display by A,B,C, ...
as well as by other concepts. )

For example, we could have the following toplevel options:
- search by word. This works already (the only problem is that if
the user doesn't hit they may have no reason why not. I am VERY
frustrated by search systems which just say 'enter search' I try a dozen
different things and get 'no hits' every time. I really need to see some
specimen entries to see what the database HAS fgot!!

- alphabetical listing. We have this as well. No problems with
this other than bulk of display.

- structured alphabetic listing (A,B,C...) this should be
straightforward as well.

- concepts. This glossary contains:
physical chemistry
geometrical terms
structural hierarchy (primary, sec...)
other :-)

The reader can then choose whether they want to look for a single term or
browse a concept. If well organised that is VERY powerful.

> Why do we need a subtree at all? Can't the main page simly be the search page -
> as I understand it that's the main use of it (this could also include an
> alphabetical search if people wanted to look through lists). As long as the
> description of all terms include a category name (i.e. all proteins have the
> word 'protein' in their description/classification) then a list of proteins
> could be found through the search.

Exactly. There is no reason why terms can't have more than one concept.
E.g. cysteine could be an 'aminoacid' and also be part of the
'disulphide' concept
> > Kurt
It's notxb as awful as I have made it sound - I think you are half way
there already! The main thing is to get good tools to support it.

> --------------------------------------------------------------------------
> Dr. Kurt Giles
> Department of Structural Biology | e-mail:
> Weizmann Institute of Science | vox: +972-8-9343759
> 76100 Rehovot, Israel | fax: +972-8-9344159
> ** **
> --------------------------------------------------------------------------

Peter Murray-Rust, Glaxo Research & Dev. (; (BioMOO: PeterMR)
Birkbeck College,, CBMT/Daresbury,