[MOBY-dev] Shims in BioMOBY

[Duncan Hull]

Hi Paul

Paul Gordon wrote:

>This is a very general question, especially for non-native-English 
>speakers.  I was going to add  a top-level "Shim" category of services 
>to the ontology,  but I'm pretty sure most non-native-English speakers 
>would have no idea what this means.  I'm borrowing the term from
>
>D Hull, R Stevens, P Lord, C Wroe, C Goble (2004). Treating shimantic 
>web syndrome with ontologies. First AKT workshop on Semantic Web 
>Services (AKT-SWS04).
>  
>
Since I'm first author of the paper [1] you mention, here are three 
points on how I think shim services should be described, so that users 
can find them. According to me, I am the first person to use the term 
"shim" to describe bioinformatics services, although its been used in 
hypermedia research [2] to describe similar operations in software.

Firstly, it is probably useful to annotate services as a shims, although 
this is information that the user  may not always want to see. A service 
that maps between equivalent identifiers (GenBank to EMBL for example) 
in one workflow might be considered a safe or "boring" operation [3], 
with shim-like properties. However, the same service in a different 
workflow constructed by a different user, might not be a considered a 
shim, because the safeness of the operation depends on the GenBank 
identifier the service takes as input. So it is probably a useful 
annotation, and a useful concept in your ontology, but the user will not 
always care wether a given service is a shim or not.

Secondly, one thing that characterises shims is the relation between the 
input and the output. So, for example, a service that extracts the 
protein sequence from a BLAST_report is a fairly boring shim, because 
the relation between the input and the output is hasPart eg. a 
BLAST_report hasPart protein_sequence. I personally think this relation 
between input and output is one that will help distinguish the shim from 
other services in the registry, and therefore aid retrieval. So 
capturing this relation in the registry could be useful.

Finally, if you're using a description logic reasoner, an interesting 
solution would be to describe the properties of shim services, and let 
the reasoner infer which services are shims based on these properties. 
This would save you some of the trouble of annotating shims in your 
registry, and would give you a clear definition of *exactly* what you 
mean when you say "shim", since the word has several different meanings....

Speaking of which, a shim [4] can also mean a transexual person: "she + 
him = shim", according to wikipedia :)

Duncan

[1] 
http://sunsite.informatik.rwth-aachen.de/Publications/CEUR-WS/Vol-122/paper1.pdf
[2] http://eprints.ecs.soton.ac.uk/768/02/html/ 
[3] http://taverna.sourceforge.net/usermanual/docs.word.html#_Toc107043031
[4] http://en.wikipedia.org/wiki/Shim

-- 
Duncan Hull
http://www.cs.man.ac.uk/~hulld/
Phone: +44 (0) 161 275 0677