I was at the 2007 UK e-Science All-Hands meeting in September 2007. This meeting is intended to cover all the e-Science work being done in the UK, but its scope has gradually expanded so that it has acquired a large international audience.

There's a lot of good stuff described here, available in the proceedings. There were two talks, however, which seemed particularly relevant to DS5.

Jenny Ure et al.: a socio-technical perspective on ontology development in healthgrids

PDF

This is a social science paper discussing the various ways that ontologies (and I suppose also vocabularies) can be developed. She described as observed possibilities having a single monolithic ontology, a small core plus extras, a number of peers with mediation between them, or an ontology created by the refinement of a folksonomy. This creates various tensions.

Alistair Miles: Collaboration in the Value Grid for Semantic Technologies

PDF

A discussion of the costs involved in creating ontologies and vocabularies, and how these can be justified. He said that he had elaborated his position somewhat between the paper deadline and the meeting, but the summary (in my notes) was:

• "Ontology development can be costly, uncertain, with poor ROI"
• You need fairly quick/cheap ontology-related gains to build trust -- deliver on investments: "realise value early and often"
• Collaboration engineering: creating an ontology is a complicated business, requiring collaboration between rather distinct parties (the standard engineer vs domain-expert issue). There wasn't much to say on this, it seemed, other than that there are carefully-thought out ways of engineering such collaboration
• Case-study: ISIS metadata (RAL neutron source). Currently lots of uncontrolled vocabulary. Describes how to add value to this in a fairly systematic way, with each stage resulting in a `product' which can be deployed.

-- NormanGray - 16 Oct 2007