On the role of linguistic descriptions of data in the building of natural language generation systems

This paper explores the current state of the task of generating easily understandable information from data for people using natural language, which is currently addressed by two independent research fields: the natural language generation field — and, more specifically, the data-to-text sub-field — and the linguistic descriptions of data field. Both approaches are explained in a detailed description which includes: i) a methodological revision of both fields including basic concepts and definitions, models and evaluation procedures; ii) the most relevant systems, use cases and real applications described in the literature. Some reflections about the current state and future trends of each field are also provided, followed by several remarks that conclude by hinting at some potential points of mutual interest and convergence between both fields.