Re: How to build an ontology, was Re: Open Terminology.

[Philippe AMELINE]

Andrew HO :

Hi Philippe,
  From your experience, do you have any suggestion regarding "how" to
build and maintain an ontology? Tim said  that it is a huge project and I
agree.

Hi,

First, I would like to explain a very important thing : in Odyssee, the ontology is not a "plus", it is the core component, and can be compared with langage for human beings.

So, even if building the ontology is a huge project, you can perfectly see Odysee as a "project around an ontology". Then Odysee is a huge project around a huge project.

I believe that you can't have accurate opinions about ontologies if you don't have a very accurate idea of what you want to use it for.
Inside Odysee, the ontology is used for any data worth sharing : I mean worth sharing between users but also between the user and the computer. It is the reason why we use a centralized approach to have the ontology evolve.

Are you sufficiently funded specifically to continue to build the
ontology? Or, are you funded to build systems that use the ontology?

We can't separate the both of it.

Approximately how many man-hours did it take to build Odysee's current
ontology?

I cannot evaluate it ; it is a monk's job. However, you must understand that in the mid 90s, I realized that unless I could build an ontology, I won't ever be able to go any further with Nautilus. We were at a point where communicating meant having a common langage.
Now, this work had the same effect for us that the compass had for sailors : working on non-localizable systems.

It is the reason why I don't totally agree with Thoma's point of view about ontologies : if you don't build a genuine common langage, the ontology is worth nothing.

I don't think that SNOMED failure is due to centralization, but came from 3 major weakness :
- starting from pathologists view angle demands a super-hughe work to reach the GP view angle
- SNOMED has always been a strange "object" : too complex to be a genuine classification, but far too simple to be a system for genuine description.
- Business plan : it is the weakest part of all components in the medical field. If the people that are doing the job don't do it for themselves, they must find customers. Who are the customers ?

Furthermore, if I can (barely) evaluate the amount of work required for building an ontology ; I can't imagine how you could have parts of ontologies built here and there fits together : it is eventually a several time harder job.

Regards,

Philippe