Inductive Inference of First-Order Models from Numeric-Symbolic Data

A factor common to statistical techniques of data analysis is the adopted representation formalism: A tabular (zerothorder) model with almost exclusively numerical features. On the contrary, several studies on machine learning concern the induction of first-order models from symbolic data, but are inadequate for continuous data. In the paper, we face the problem of handling both numerical and symbolic data in first-order models. distinguishing the moment of model generation from examples (induction) from the moment of model recognition by means of a flexible. probabilistic subsumption test. We demonstrate the proposed solutions on a problem in document understanding. where the objective is to induce the models of the logical structure of some real business letters.