Re: BioRDF Brainstorming

Colin Batchelor wrote:
>>>Surely this is better suited to OBI-type annotations than generic OBO 
>>>things?

Please describe further how "OBI-type" differs from "OBO-type". OBI is part 
of OBO.
Do you mean quantitative information, such as detailed numerical results of 
experiments?

>>>OBI-type annotations are significantly harder to pull out of text than 
>>>the sorts of thing you mention on the wiki page."

Using NLP to accurately match entities from ontologies like GO is already 
hard enough for simple phrase recognition. Extracting descriptions of 
processes (e.g. 'there is a binding process with the participants 5-HT2a and 
Serotonin) is even harder/impossible for most currently available NLP 
software.  If we could popularize the annotation of biomedical texts and 
database entries in this way, we would already have achieved a lot. I think 
at the moment it is preferable to start with simple annotations based on 
widely accepted ontologies. Once we can still aim to create more elaborate 
annotations when we have established a good userbase and have been accepted 
by the mainstream biomedical research community.

I also think that the machine-readable representation of facts about biology 
should have a higher priortiy than the description of experimental setups 
and procedures (which is the major goal of OBI and EXPO). People only have 
limited time and motivation to create machine-readable annotations, and it 
is much more useful when they spend that time on describing the RESULTS 
(biological facts). Of course, descriptions of experiments are also 
valuable, when there are sufficient resources left for creating them.

>>>I should mention that we (RSC) are supporting 
>>>http://www.aber.ac.uk/compsci/Research/bio/art/ which aims to do just 
>>>that sort of thing.

Yes, I know about ART, and it looks quite promising. I am quite sure that 
the software and user interfaces developed by ART could also be very useful 
for OBO ontologies. Are there already some preliminary results, prototypical 
user interfaces etc.?

I am also aware of RSC's "Project Prospect", which also seems to be of 
interest in this context:
http://www.rsc.org/Publishing/Journals/ProjectProspect/

Cheers,
Matthias Samwald
Semantic Web Company / DERI Galway / Yale Center for Medical Informatics

Received on Monday, 11 February 2008 15:57:18 UTC