Stream processing is a general information processing paradigm with different applications in AI. Most stream languages rely on the concept of a sliding window with a bag semantics, which is in order for relational streams but may lead to inconsistencies when applied on streams of assertions evaluated against a conceptual model. Our approach uses a different semantics based on ABox sequencing. The query language provides an expressive first order temporal logic for inter-ABox reasoning. Safety conditions tame the expressiveness so that a meaning preserving transformation of the query to backend queries on the sources as foreseen in the OBDA paradigm is guaranteed [1].
PREFIX : <http://www.sport.com/Fitness/Pattern#>
CREATE PULSE pulseA WITH
START = "2012-07-04T06:21:58CEST"^^XSD:DATETIME,
END = "2012-07-23T00:54:35CEST"^^XSD:DATETIME,
FREQUENCY = "PT1M"^^XSD:DURATION
CREATE STREAM S_out AS
CONSTRUCT GRAPH NOW { ?profile a :IncreasingHeight }
FROM STREAM measurements [ NOW - "PT100M"^^XSD:DURATION, NOW ] -> "PT5M"^^XSD:DURATION
USING PULSE pulseA
SEQUENCE BY StdSeq AS SEQ1
HAVING FORALL i, j IN SEQ1 ?x,?y(
IF ((
GRAPH i { ?profile :hasHeight ?x }
AND GRAPH j { ?profile :hasHeight ?y }
) AND i < j) THEN ?x <= ?y)
