Generating Linguistic Descriptions of Data Using Fuzzy Set Theory 1

In this extended abstract we give our view, as introduced in [1], of the problem of generating linguistic descriptions of data (GLiDD). This task is performed by data-to-text systems, which have their origin and more advanced development in the field of natural language generation (NLG). These systems aim at generating texts, we call linguistic descriptions of data, from non-linguistic input data, expressing knowledge extracted from the latter as humans would do, with the objective of satisfying specific user’s needs. As different authors have shown, using a brief text is a feasible, and sometimes the most effective, mean for data description. This is particularly the case when the devices employed are based on written or spoken natural language, when the user is visually impaired, when the understanding of visual or mathematical means requires expert knowledge or complex cognitive tasks, etc. [1]