Linguistic Summarization Based on the Inherent Semantics of Linguistic Words

Abstract

Linguistic summarization provides an explicit, concise and easily understandable description expressed in natural language of huge amount of data for human. The result of extracting linguistic summaries is natural language sentences using given sets of words for each numeric attribute. In general, the word-set is always assumed to be of only \(7 \pm 2\) words represented by fuzzy sets. The interpretability of linguistic summarization depends mainly on this fuzzy set representation of the words. In this paper, we consider the given word-set for each attribute as a linguistic frame of cognition LFoC of this attribute whose size depends only on application requirement. We propose a linguistic summary method based on multi-granularity representations of this LFoCs that can preserve order-based semantics relation and generality-specificity relation of words. Theoretically, the number of words in LFoC are not limited. A simulation study using dataset Iris shows that the proposed method can extract sentences using words of length 3 characterizing dataset Iris that other existing ones cannot do.