| Resumen |
Can we retrieve objects from a database that are just close to a desired item? Can we control the retrieval process up to a given error? Can we define a semantic similarity between objects to be retrieved? The paper seeks for positive answers to these questions by using hierarchies. Sometimes, qualitative values can be arranged in a hierarchy, a tree with certain properties. Hierarchies allow the definitions of following constructs: the confusion conf(r, s) in using qualitative value r instead of the intended or correct value s (such as using ‘umbrella’ when ‘raincoat’ was intended); a predicate P that holds for object o within confusion H, written PH(o); the closeness of an object to a predicate (closeness of fit); how close or
similar two objects are; predicate PH (o), where the total (cumulative) error produced by object o is at most H. These constructs are used in an extended SQL notation to retrieve, from a database, answers imperfectly fulfilling a predicate, up to a desired confusion (error). The extended SQL expression (a query) is automatically transformed to normal SQL, which then retrieves. It is shown how to extend any relational database for precision-controlled retrieval and updates. Examples are given. |
