Example-based automatic induction of semantic classes through entropic scores

Abstract - The paper deals in some detail with the application of examplebased machine learning techniques to the task of automatically acquiring semantic information from functionally annotated texts. Special emphasis is placed on the use of “analogical proportions” as a means of structuring the knowledge embodied in attested examples, and weighing up their contribution to a variety of lexico-semantic classification tasks. Careful quantitative analysis of automatically acquired information proves to shed considerable light on the semantic inter-connectivity of input data, their structure and organising principles.