RDFIG/WebOnt notes ----- Trust modelling ... originally came up in CC/PP, re handling of conflicting capability descriptions ... subsequently considered as an extension of PKI technology for authorization ... along the way met up with folks at RAL ... CORAS - http://www.nr.no/coras/ iTrust - no web page yet ------ ... Q for Jos: proofs with "variables" (c.f. compiling) - YES; dereferencing URIs; fwd vs backward ... Martyn -- "intellectual content" forced in response to customer wishes -- calls for flexibility of RDF? ... Survey of query languages ------- Conceptual graphs (Matto Acacia? Inria) CORESE Application: corporate memory CGs have only one type per relation. CGs have strict separation between concepts and relations + Additional semantics --> have mapping to DAML + Simple rule engine + No existential nodes in rule conclusions + Simple RDF syntax (pattern with variables) + Operators within literals (contains, ...) Query by projection (find specializations of query that match the data ?) Uses CG inference engine, with RDF pre/post conversion. Demonstration: uses schema relationships to add semantic element to search??? www.inria.fr/acacia, in Java ----- jjc: Inference and querying -- build inferencing on simple query model, or a more generic (e.g. SQL-like) query. I have concern about inconsistent/unpredictable results if the inference rasoner is not complete -- DanC made the point that DL folks place high importance on complete reasoner. Union of graphs (query over logically merged graph) -- maybe have different constructors that do or do not support write-back, with different choices about where to write back. ----- MikeD - integrating data sources JDK HTML parser delievers clean event stream. Then special code to scrape RDF info. Ugly code, but only 100 lines of ugly; e.g. java program to scrape info out of NYSE stock code pages, etc... Similar add-in for cwm from Mark Nottingham. Currently, lot of disconnected islands. Definitive sources for common information would help here. Processed 80% of info from CIA world factbook, to RDF. (MikeD's work uses early binding; cwm uses late binding) ***** can markov chains be used for this kind of information scraping ****** ***** (c.f. voice recognition) ****** "These things are incredibly fragile" Suggested: a screen-scraping _service_ but also remember schema info provided (a variation on Google???) Use this in conjunction with automated schema ... MikeD: focused too much on the schema-definition parts of this stuff, when what most people want to do is trivial (in terms of classes/properties) ----- MikeD on DAML (or WebONT?) Charter is good summary of semantic web issues. To be known as OWL. -----