Natural language-based automatic qualitative description of clusters

In this paper the design of a natural language generation (NLG) system is introduced to describe qualitatively the most important characteristics of each class or cluster previously defined by means of a classification or clustering process. An adaptation of a generic architecture for data-to-text systems consisting of four stages is proposed. It includes the detection of the most relevant patterns of the data and the definition of a grammar that generates the natural language description of the considered clusters. A case study addressing a challenge in a marketing environment is included. The study takes place in a business-to-business (B2B) environment, in which a firm distributes its products via other firms.