> On 26 Nov 2018, at 15:26, William Waites <wwaites@tardis.ed.ac.uk> wrote: > > >> honestly, if those guys [1] get by without literals as subject, i think >> everybody should :-) > > Hah! Sure, but there’s a deeper problem that this exposes. It’s the absence of > any real way to specify context for reasoning. It’s the reason why it’s normal to > ignore the meaning of owl:sameAs or simply avoid using it. If you take all the > triples and try to apply all the rules, you get nonsense. Having literals as subject > is not the source of the nonsense. The nonsense is caused by having no standard > way to do reasoning while keeping the results under control. Some related food for thought: Named graphs as containers in respect to contexts and departing from the open world assumption. Some things are only true in a particular context whilst other things are deemed to be generally true. Data schema may describe closed enumerations so that if you know something has a given value, you know it doesn’t have another value from the given enumeration. Similar considerations apply to defaults where something has a particular value rather than the expected default. Path expressions for context sensitive data mapping rules. These could for instance make use of ATNs for graphical rule languages. Essentially things may only be equivalent in particular contexts rather than generally. You could consider this by analogy to translating between human languages. Blending ideas from cognitive science and computational statistics for reasoning and recall in terms of rational belief that is based upon prior knowledge and past experience, and capable of addressing uncertain, incomplete and inconsistent knowledge and data. This takes an inter-disciplinary approach, and could build on top of the RDF extensions we have been discussing on this list. In case this sounds too abstract - web search is a concrete and highly successful proof of concept for combining symbolic representations with statistical information and its application to personalised search results. Dave Raggett <dsr@w3.org> http://www.w3.org/People/Raggett W3C Data Activity Lead & W3C champion for the Web of things
Received on Monday, 26 November 2018 15:56:20 UTC