On Implementing Temporal Query Answering in DL-Lite (extended abstract).

Temporal information plays a central role in many applications of ontology-based data access (OBDA). For example, knowledge about the past is usually kept in patient records, and collected by companies or scientific projects as MesoWest4, focusing on weather data. Such applications obviously benefit from using ontologies for data integration and access (e.g., the wind force ‘Storm’ on the well-known Beaufort Wind Force Scale is equally characterized by wind speed and wave height, which can be represented by a general concept inclusion as HighWindSpeed t HighWaves v Storm). Temporal knowledge is however not taken into account by systems implementing OBDA, in general. Though, assuming that we consider several weather stations’ data of the past 24 hours, a query such as the following could be interesting: “Get the heritage sites that are nearby a weather station, for which at some time in the past (24 hours) a danger of a hurricane was detected, since then, the wind force has been continuously very high, and it increased considerably during the two latest times of observation.” For that reason, we investigate different approaches for answering temporal conjunctive queries (TCQs) [4, 5] w.r.t. ontologies written in the description logic (DL) DL-Litecore (hereinafter called DL-Lite). TCQs combine conjunctive queries (CQs) via LTL operators5 and have already been studied extensively in the context of DL-Lite [13, 8]. The above example query could be specified as the following TCQ: HeritageSite(x) ∧WeatherStation(y) ∧ nearby(x, y) ∧ ( HighWind(y)SDangerOfHurricane(y) ) ∧ #− Storm(y) ∧ ViolentStorm(y), asking for all pairs (x, y) of heritage sites and nearby weather stations, whose sensor values at some point in time implied a danger of a hurricane, since (S) then, the measurements have implied Beaufort category ‘high wind’, in the previous (#−) moment of observation they implied category ‘storm’, and the latest values imply ‘violent storm’. The semantics of TCQs is based on temporal knowledge bases, which, in addition to the ontology (assumed to hold at every point in time), contain a sequence of fact bases A0,A1, . . . ,An, representing the data collected at specific points in time. Especially note that the ontology and the fact bases itself are formulated in a classical DL.