ESWC2007

Norman was at the European Semantic Web Conference, 2007

The tutorials beforehand, and the workshops afterwards, turned out to be substantially more interesting and useful than the two-day meeting itself. Of those, the tutorial on uncertainty reasoning and the scripting workshop seem the most relevant to the VO, so they're the ones I'll discuss here, with mentions of posters elsewhere in the meeting.

Tutorial: Uncertainty and Vagueness

The two notions are different: `uncertainty' refers to statements which are true or false, though we cannot be certain which; that includes statements such as this is a spiral galaxy. Vagueness refers to statements which are not straightforwardly true or false, such as the weather is hot today or this is a useful VOResource. Though these notions are quite distinct, the underlying formalism for dealing with them turns out to be the same. In both cases, folk talk about a degree of truth between 0 and 1.

The idea of uncertainty reasoning is to develop a calculus combining values such as this, in such a way that it mimics traditional logic: if X is probably a bird, and most birds can fly, can X fly? One way of doing this, of course, is to use Bayes theorem a lot, but there are other schemes, with various attractive and unattractive properties, suitable for different logical applications. For example, not all implementations have not(not(X)) = X, and not all applications necessarily need that.

There are multiple implementations for uncertain logic programming, none for uncertain description logics, and I've got `yes?' written down next to `uncertain RDF', so heaven knows what that means.

Workshop: scripting on the semantic web

A recurrent theme of this workshop was

• There's lots of semantic/structured data on the web, but it's not necessarily marked up as RDF.

For example, Amazon's API provides a lot of interesting structured data, and you can start to use this in SemWebby ways as soon as you put something in front of it to transform it into triples.

This is a good point to mention the various ways of doing that transformation:

• RDFa is a way of embedding triples (or more generally semantic information) within HTML using a standardised vocabulary
• GRDDDL is a way of embedding triples in X(HT)ML in a very general way
• Microformats are fashionable right now (I think: they might just be soooo last-month)
• D2R is a server that can sit beside a RDBMS, so that it translates SPARQL queries to SQL on the fly, effectively exposing non-RDF databases as virtual RDF graphs (unlike the approaches above, this is completely hands-off, and doesn't require changes to the information being served)
• Not generic, but DBpedia is a way of making semantic queries of the information in Wikipedia; it has the same motivation, but a different approach, as Semantic MediaWiki
• Create a custom service. The RDF Book Mashup takes an ISBN and returns RDF information about it using various occult means, including using the Amazon API.

Not in that workshop, but on the same theme, there was a poster about SemDAV, which is a (currently proprietary) extension to WebDAV which is intended to be a `semantic filesystem'. The idea seems to be that, instead of the simple hierarchical structure of a directory-based filesystem, you can have just the equivalent of `smart folders', which automatically group filesystem-type objects based on their metadata. It sounded like fun, but perhaps not immediately deployable right now....

There were multiple mentions of what seem to be the consensus technologies for browsing RDF data: Fresnel, Disco and Watson appear to be the key words here, and playing with them is on my list.

Thinking and talking about it, I feel that an RDF Registry (coming soon!), D2R services in front of some important astronomical databases, perhaps DBPedia too, why not, all mixed together with a bit of imagination, could make some very interesting user-facing applications. That little lot wouldn't be trivial to implement, but I don't believe it would be an impossibly huge investment, especially since the whole architecture is built around the various components being independent, and separately mashupable.

-- NormanGray - 2007 July 23