Extracting Composite Summaries from Qualitative Data

Abstract

The paper proposes a model combining association rules and elements of Rhetorical Structure Theory (RST) to generate composite linguistic summaries from qualitative data. The specifications of three new abstract forms of composite linguistic summaries for qualitative data are presented. The proposed abstract forms represent relations of Evidence, Contrast, and Emphasis inspired by RST, consisting of at least two semantically related constituent statements linked by a connector specific to each relation. The constituent statements have the structure of the classical protoforms of linguistic summaries, which in this paper are built from an association rule, to which a fuzzy linguistic quantifier is assigned. Moreover, the definitions of truth degree, relation strength, and coverage degree for composite relations are presented. The model applicability was checked through a use case performed with a database of 2128 cases of the Economic Chamber of the Provincial People’s Court of Havana.