Saturation: A data transformation within a format and structure that only adds data derivable (according to some semantics) or otherwise computable from the source data
This is sometimes also called derivation, however in IN5800 we have used this only with respect to reasoning and semantics. But saturation is more than just semantic derivation, e.g.:
- Value extraction: Extracting values from within more complex values, such as extracting the city name from an address, and then making the relationship between the building and the city it is in explicit
- Value translation: Translate a value from e.g. one unit of measure
to another (feet to meters), or e.g. translate from quantitaive to
qualitative representations (
- Computing attributes/relationships from complex objects: Many datatypes contain lots of information within them, such as time intervals or spatial objects. Examples of this type of saturation is computing spatial relationships such as overlaps, touching, containment, etc. between geometrical or interval objects. Examples of attributes that can be computed from such objects are duration/area/extent, dimensionality, conectedness of temporal or geometrical objects.
- Aggregation/Object creation: Combining multiple values into one. E.g. one can combine the locations (points) of multiple sensor values measuring a wind speed greater than 20 m/s into a single object denoting a storm.
- Interpolation/extrapolation: Adding data points within (interpolation) or outside (extrapolation) of the range of a current set of datapoints. E.g. one could interpolate a weather forecast containing an hourly forecast to a forecast for every half hour.