Conceptual defaults in fuzzy ontology

The paper explores the fuzzy status of implicit information in natural language text, focusing on conceptual defaults, the routinely omitted information that readers/hearers equally routinely reconstruct. Making this information available to the natural language processing computer is essential, and fuzziness is a major issue. An analysis of 1,000 English sentences has demonstrated a diverse combination of circumstances for detecting and computing defaults with their membership function values.

I. Introduction

There is no doubt that information presented on the web, newspapers, or any other media is imprecise, and we preach and practice the fuzzy-logic treatment of this imprecision. There is also no doubt that we rely on some details that we assume are recoverable from the text even when we are not explicitly stating them—in fact, it is easy to demonstrate [1], Ch. 6 that every sentence carries much more information for the hearer/reader that it explicitly states, and that this informations comes with varying membership function values with regard to its epistemological modality, i.e., whether it actually happened.