In the context of collecting disparate data sets holding similar information, are their examples of algorithms being able to resolve attributes of records being similar (while their values are different) based on their relation to other attributes? (So fuzzy matching would not necessarily apply)
Simplistic example: 2 Datasets regarding vehicle test data (perhaps with headings in different languages)
- Source A: Metric information: Car, Distance tested (km), Fuel Used (l), Consumption (km/l)
- Source B: Imperial information: Car, Distance tested (miles), Fuel Used (gallons), Consumption (mpg)
- Mapping between A & B, which recognises that the consumption data in different units, is actually the same because they are simply "column 2 / column 3" ?
- Could it identify that the distance columns were holding similar information with a 1.6x scaling factor applied etc?
- In this hypothetical example, matching could apply on column 1 to identify cars for record linkage, however I'm curious to see if there are methods which can identify the relationships between attributes, and recognise where attributes are similarly linked in other sets.
The closest material I have found seems to be this paper however I'm under the impression that this is probably a well investigated problem where I am not aware of the correct avenue to search!
Thanks for your input!