On 29 May 2013 11:21, Adrian Giurca <giurca@tu-cottbus.de> wrote: > Dear Bernard, > Indeed extracting temporal information from text is a complex task. As > such, if you consider automatic schema.org data creation, then of course, > there are many challenges. However, looking from the perspective of human > and machine data creation, i.e. aiming to provide a data format for Opening > Hours then I agree with Good Relations because makes possible various > combinations. Many examples discussed in paper [3] (e.g, "open from > February 15th to 30th", "Open every day in July and August. Outside this > period, closed on Wednesday night, all day Thursday and on Sunday night. > Opening times: from noon to 2pm from 7pm to 10pm.") can be markup with > OpeningHoursSpecification, of course, sometime easy, sometime much more > verbose. > > Just to try an answer to Richard: I believe in this case (and possibly > others) the meaning of missing information should be interpreted as > negative information (Closed World Assumption). As such when, for example, > Wednesday is completely missing from markup, then the machine means > "closed". > If we will adopt the Open World Approach then the negative information > must be explicitly expressed i.e., we need to encode "closed" explicitely > therefore a proposal such the one Bernard did below is suitable. > (Quick aside from an airport) This is actually one of the places where microsyntax (stuffing lots of information into a single string) versus using several fields matters - to respond also to Matthias' point earlier in this thread. Schema.org data is open world in general, consisting of collections of simple factual statements (triples). And in general you can't assume anything from the absence of some of these statements/triples. However when information is expressed all within a single structured string value, it gets safer to make closed world assumptions. Dan
Received on Wednesday, 29 May 2013 10:54:56 UTC