OBSERVE Semantic Layer

This semantic layer (http://semanticweb.cs.vu.nl/observe) contains vocabularies and proof-of-concept for the RAAK-mkb Observe project that accompanies a paper currently under submission. It is realized using Cliopatria, the SWI-Prolog Semantic Web Server. More information can be found at the Github page: https://github.com/aan680/Observe.

Schema and Vocabularies

The schema of observe and that of the vocabularies it depends on are stored in separate graphs:

http://semanticweb.cs.vu.nl/observe/browse/list_graphs.

Demonstration Scripts

The demonstration scripts (in Prolog) are on the server too:

http://semanticweb.cs.vu.nl/observe/swish/p/Spreadsheet_to_observations.pl http://semanticweb.cs.vu.nl/observe/swish/p/Define_context.pl http://semanticweb.cs.vu.nl/observe/swish/p/Context_to_source_observations.pl http://semanticweb.cs.vu.nl/observe/swish/p/Collect_context_profile.pl http://semanticweb.cs.vu.nl/observe/swish/p/Open_data_for_context_profile.pl http://semanticweb.cs.vu.nl/observe/swish/p/Content_matching.pl.

If the (collaborative) SWI-Prolog editor appeals to you, have a look at http://swish.swi-prolog.org.

Documentation for the above scripts:

Process spreadheets with observations:

• Access: http://semanticweb.cs.vu.nl/observe/swish/p/Spreadsheet_to_observations.pl
• Input: no input, but the script expects a spreadsheet ‘data_demo_observe.csv’ uploaded to the server (only admin can do this) with rows like this: Enschede,52.2167,6.9,30,2015-07-15T09:00:00,20
• Output: RDF triples in Data Cube, i.e. defined qb:Observation instances.
• Syntax: main.
• Example: main.
• Future work: Make the methods more generic. All properties are currently pre-defined (temperature, date time, location, lat/long, etc), so methods could be defined to declare attributes with any name and with values of any kind.

Define contexts:

• Access: http://semanticweb.cs.vu.nl/observe/swish/p/Define_context.pl
• Input: details about the context, i.e. a time (in sxd:dateTime notation) and location (in lat and long) corresponding to the screen position.
• Output: RDF triples defining the context (with a URI as ) and the properties refTime and refArea
• Dependencies, prerequisites: none.
• Syntax: define_context(dateTime, lat, long).
• Example: define_context('2015-07-15T09:00:00', 50.000, 5.000).??

Connect contexts to source observations: ?

• Access: http://semanticweb.cs.vu.nl/observe/swish/p/Context_to_source_observations.pl
• Input: maximum allowed time difference in seconds between observations and contexts for the former to be relevant to the context.
• Dependencies, prerequisites: requires defined instances of dq:Observation and observe:Context
• Output: RDF triples relating all ob:Context instances to all dq:Observation instances with the same lat/long and a difference in refTime of at most [MaxDelayInSeconds], through predicate observe:source
• Syntax: connect_contexts_to_sources(MaxDelayInSeconds).
• Example: connect_contexts_to_sources(300).
• Future work: Currently the observation and the context need the same lat/long as their reference area.

Collect the (first part of the) context profile based on observations:?

• Access: http://semanticweb.cs.vu.nl/observe/swish/p/Collect_context_profile.pl
• Input: none, although the script depends on (pre-defined) context attribute-value definitions, consisting of the name of the ContextAttributeValue (e.g. sunshine-high), the name of the measurement it refers to (e.g., sunshine), and the lower and upper bound float numbers that are the criteria for the measurement to get this context attribute-value (e.g., between 0 and 0.50).
• Output: RDF triples with all the properties of a given context. Method all_contextprofiles fills out the contextattribute values for all contexts defined, based on the Observation instances that are detected in the previous step as a valid source for the context profile.
• Syntax: all_contextprofiles. This collects the context profiles. define_contextattributevalue(Name, PropertyName, Lower, Upper). To define a new context attribute-value pair, if needed.
• Examples:
  all_contextprofiles. define_contextattributevalue('crowdedness-low', 'crowdedness', 0, 0.09).?
• Future work: define the context attribute values outside the Prolog environment, for instance via a comma-separated file in the same folder as the source observations.

Add to the context profile data from the web of Open Data:

• Access: http://linkedpolitics.eculture3.labs.vu.nl/swish/p/Contextprofile_from_open_data.pl
• Input: a type of venue the occurrence of which is queried, as a Linked Geodata category; a lat and long coordinate around which it is queried; a radius in metres around which to look for such venues
• Output: the presence (true/false) of venues of the specified type within the square region that spans the specified number of metres around the specified lat/long coordinate, if desired also the name of each venue. Add a 'count' operation to the query from the 'solutions' tab to retrieve the number of venues instead.
• Syntax: venues_of_type(VenueType, Lat, Long, DistanceInMetres). venues_of_type(VenueType, Lat, Long, DistanceInM, VenueName).
• Example: nr_of_venues_of_type('http://linkedgeodata.org/ontology/School', 6.917, 52.210, 500).
• Future work: 1) make qualitative context profiles based on ratios of different types of venues. For instance, mark an area with more museums than schools as a tourist area. 2) consider downloading the relevant set of the Linked Geodata database for reliability.

Match content items to contexts:

• Access: http://semanticweb.cs.vu.nl/observe/swish/p/Content_matching.pl
• Input: none
• Prerequisites: objects of type observe:ContentItem with ob:suitable-for relations in the RDF store.
• Output: RDF triples relating all ob:Context instances to all ob:ContentItem instances that are suitable for that context, through property ob:suitable-for. By ‘suitable for a context’, we mean suitable for all the attribute-value pairs defined for the context.
• Syntax: match_all_contexts.
• Example: match_all_contexts.